Speed!

[mikelbeck]

I've got a question about speed to my VPS...

It seems that when I first connect to a site on my VPS, it takes a while before the page comes up. After I'm connected, it's pretty quick, it's just the initial connection.

I did a tracert on dnsstuff.com, and it takes 28ms to get to my VPS. The old host I was on, it took 3.2 ms to get to the host. To get to powervps.com, it takes 2 ms.

Is this something in the configuration of my VPS, or something in the network?

Here's the trace to powervps.com: http://www.dnsstuff.com/tools/tracert.ch?ip=www.powervps.com
Here's the trace to a site I still have active on my old host: http://www.dnsstuff.com/tools/tracert.ch?ip=www.elite-computing.net
And here's the trace to my VPS: http://www.dnsstuff.com/tools/tracert.ch?ip=www.ecshosting.net

Once it gets past the firewall that doesn't respond, all of the responses are 28ms. Why is this, and is there anything that can be done about it?

[charles]

Hi

The initial delay is something we have seen for idle sites. If the site has been idle it seems it takes longer for the very first attempt due to an arp lookup needed (since it's expired). This is due to the Virtuosso technology.

The difference in traceroutes is because powervps.com is on a different bandwidth mix (single homed on *** like servint ) whereas your VPS is on our BGP mix of providers. Your path to your vps tkes a different route (through nlayer in this case).

I dont really trust that output with that router. I suggest running mtr (or winmtr) from you desktop to your vps, and at the same time run mtr on your vps back to your desktop ip. It will show you the true routes and times for each hops as we use asymetric routing.

28ms is not bad at all though.

hth
charles

[mikelbeck]

The initial delay is something we have seen for idle sites. If the site has been idle it seems it takes longer for the very first attempt due to an arp lookup needed (since it's expired). This is due to the Virtuosso technology.

Is there anything that can be done about this? Some tuning, enable some parameter somewhere or something?

28ms is not bad at all though.

No, it's not, but with my old host I was getting 3ms, I thought that had something to do with the speed of accessing my VPS.

[charles]

Is there anything that can be done about this? Some tuning, enable some parameter somewhere or something?

No, there is nothing you can do. We are working with swsoft on this but don't have a resolution yet.

charles

[mikelbeck]

Ok, here's what I'm seeing:

- The first visit to a site on my VPS takes a long time to connect. I guess this is the issue you responded to above.
- The first connect via POP3 to a site on my VPS takes a long time to connect... 10, 15 seconds some times. Is this the same issue?
- WHM & CPanel both do the same thing. When I first connect to a domain it takes a long time to connect, after I'm connected it moves along a little (but not much) faster.
- SSH does the same thing. Slow to connect, then while I'm connected it speeds up, but every once in a while it grinds to a halt.

CPU, memory isn't loaded while this is happening, so I don't think it's a machine issue. Could it all be network related?

What kind of arp lookup is it trying to do? There's nothing in the arp cache on the VPS when I'm connected via SSH and logged onto a site on the VPS.

[charles]

Please run mtr on your pc to your vps. You can also run it on your vps to your pc (or ask us to). This will show if you have any network issues. Send support the results both when its good and when you are seeing a problem.

The arp cache issue isn't on your VPS. Rember this is a virtual private server and all traffic effectively goes through the physical hardware node (shows up as a router hop).

[Hvu]

Weird, I have never had this happen to me. SSH / web everything loads quick. Could it be DNS settings? Before using named on my server it was a bit slow, Now i'm using dnsmadeeasy.com and its just lightening fast.

[mikelbeck]

Weird, I have never had this happen to me. SSH / web everything loads quick. Could it be DNS settings? Before using named on my server it was a bit slow, Now i'm using dnsmadeeasy.com and its just lightening fast.

You're using dnsmadeeasy for your primaries? I use them for my secondaries, and used named as my primaries.

Charles - should I post the results in this thread, PM them or open a ticket?

[Hvu]

DNSmadeeasy takes care of all my DNS needs. Named doesnt get used on my server.

[mikelbeck]

DNSmadeeasy takes care of all my DNS needs. Named doesnt get used on my server.

What's the advantage to doing it that way?

[Hvu]

Well, Not sure since you use named and DNSmadeeasy. Its about the same thing as using DNSmadeeasy only.

[charles]

Please open a ticket. It's almost impossible to support you via these forums and we have perfectly good tools that save my little brain from having to remember you, your vps(s) previous tickets etc. etc. etc.

charles

[mikelbeck]

Please open a ticket. It's almost impossible to support you via these forums and we have perfectly good tools that save my little brain from having to remember you, your vps(s) previous tickets etc. etc. etc.

charles

Ok, I'll leave your brain alone. ;-)

Tomorrow morning I'll open a ticket.

[mikelbeck]

Please open a ticket. It's almost impossible to support you via these forums and we have perfectly good tools that save my little brain from having to remember you, your vps(s) previous tickets etc. etc. etc.

charles

I submitted a ticket this morning.

[shoopie]

I've got a question about speed to my VPS...

It seems that when I first connect to a site on my VPS, it takes a while before the page comes up. After I'm connected, it's pretty quick, it's just the initial connection.

I did a tracert on dnsstuff.com, and it takes 28ms to get to my VPS. The old host I was on, it took 3.2 ms to get to the host. To get to powervps.com, it takes 2 ms.

Is this something in the configuration of my VPS, or something in the network?

Here's the trace to powervps.com: http://www.dnsstuff.com/tools/tracert.ch?ip=www.powervps.com
Here's the trace to a site I still have active on my old host: http://www.dnsstuff.com/tools/tracert.ch?ip=www.elite-computing.net
And here's the trace to my VPS: http://www.dnsstuff.com/tools/tracert.ch?ip=www.ecshosting.net

Once it gets past the firewall that doesn't respond, all of the responses are 28ms. Why is this, and is there anything that can be done about it?


I was about to ask the same question just now... i also did the exact same thing you did at dnsstuff.com (webmasters instinct) :D

[mikelbeck]

I was about to ask the same question just now... i also did the exact same thing you did at dnsstuff.com (webmasters instinct) :D

Are you having the same issue that I am?

[smoore]

I've only seen a pause two or three times in the last couple months.

Perhaps this idle timer isn't as much of a factor if you have fairly steady traffic to the VPS? I wonder what the timeout period is, and if something like a website monitoring tool might help keep the arp entry loaded.

[TomK]

I think we have a few issues here, that we are working on:

1 - BGP via nLayer and HOP1 in Ashburn, as HOP1 is sending to nLayer in Chicago then we back-haul it to Ashburn, VA - shouldn't have to do this, we are working on changing the BGP routes now.

2 - ARP issues/stale ARPs, we have a working ticket with SWsoft to see if the ARP issue for any VPS customers that don't have a regular flow of traffic are causing a slight delay for the first packet, we are waiting to hear back from them now.

and...

3 - The initial node might be having some IO slow-downs as backups are running we are working on this now, and moving the VPS to test it.

Of interest to me are #1 (which is getting fixed now), and #2, as we've seen this on a few customers out of 1000's over the last 30 to 60 days.

I'll let you know what we find out!

[mikelbeck]

I think we have a few issues here, that we are working on:

1 - BGP via nLayer and HOP1 in Ashburn, as HOP1 is sending to nLayer in Chicago then we back-haul it to Ashburn, VA - shouldn't have to do this, we are working on changing the BGP routes now.

2 - ARP issues/stale ARPs, we have a working ticket with SWsoft to see if the ARP issue for any VPS customers that don't have a regular flow of traffic are causing a slight delay for the first packet, we are waiting to hear back from them now.

and...

3 - The initial node might be having some IO slow-downs as backups are running we are working on this now, and moving the VPS to test it.

Of interest to me are #1 (which is getting fixed now), and #2, as we've seen this on a few customers out of 1000's over the last 30 to 60 days.

I'll let you know what we find out!

Thanks, Tom, for getting involved with this. I'd appreciate it if you would keep me informed of the progress of items #1 and #2.

[TomK]

Option #1 has been resolved, I just personally did some testing and it is under 3ms now.

Option #2 has been submitted to SWsoft, we'll see what happens.

BTW, please check your ticket status, and respond in there to Rob if you can.

Thanks!

[mikelbeck]

Option #1 has been resolved, I just personally did some testing and it is under 3ms now.

Good deal. I'm still seeing higher numbers (between 14 and 20ms), but that's better than it was. mtr is reporting slightly lower numbers as well. I'll be keeping an eye on it and will let you know what I see.

Option #2 has been submitted to SWsoft, we'll see what happens.

Has anybody else reported this sort of thing to them or is this the first they've heard of it?

BTW, please check your ticket status, and respond in there to Rob if you can.

Already done. I told him the same thing, I'd keep an eye on it and would report back later on today.

[smoore]

Very nice, option #1 made a significant improvement on my VPS tracert between now and when I tried it earlier today, from around 28ms down to 2.3ms (using dnsstuff). Thanks for posting these details for those of us not involved in the ticket!

[shoopie]

Are you having the same issue that I am?
I did but now my site is loading really fast... wow ;)

looks like they fixed the problem... good job!

[mikelbeck]

I did but now my site is loading really fast... wow ;)

looks like they fixed the problem... good job!

Is it still "really fast"? Mine was very quick this afternoon, now it seems to have slowed down a bit. I'm gonna run a trace again to see if the numbers are any different.

[ndndixie]

Mine is zooming too. The NFL site that I run is twice as fast as it normally is. I just checked my one site that's usually slow and it's loading quick as well.

[ndndixie]

Ping #1: 2.9ms [TTL=59]
Ping #2: 2.8ms [TTL=59]
Ping #3: 3.0ms [TTL=59]
Ping #4: 2.6ms [TTL=59]
dnsstuff.com

[shoopie]

Is it still "really fast"? Mine was very quick this afternoon, now it seems to have slowed down a bit. I'm gonna run a trace again to see if the numbers are any different.
yup its fast... it was 28 ms earlier but now its 2.1 ms :D

[mikelbeck]

Ok, it's probably my network here then... From dnsstuff:

Ping #1: Got reply in 2.8ms [TTL=59]
Ping #2: Got reply in 2.8ms [TTL=59]
Ping #3: Got reply in 2.7ms [TTL=59]
Ping #4: Got reply in 2.8ms [TTL=59]

:D

[danweber]

Very interesting thread. I have kept support busy with many tickets over the past few weeks with slow VPS performance. I also see POP3 performance that is in the 20-30 seconds initially. Control panel (Plesk in my case) is very slow too. I believe in my case it all comes down to the fact that I don't have a steady traffic to my sites and some sort of idle timer kicks in very quickly. I even upgraded from a Power-1 to a Power-2 configuration but it did not make a difference.

I am not contemplating in getting a dedicated server instead. Why I have not much traffic to my sites I had received complains about the initial delay and I need to get this resolved. Support has been extremely cooperative and helpful in isolating the issues but not to my satisfaction (yet).

I'll keep monitoring this thread as it seems there might be some improvements ahead and at the same time I'll investigate the dedicated server route now.

Dan

[TomK]

Dan, I think your smtp/pop is port 113 identd problem, are you behind a firewall (the machine checking mail)?

Support should have asked you this in one of your earlier tickets.

The ARP issue isn't wide spread and doesn't really cause a 'slow loading' page, it causes the initial request to be a few ms in delay, but the page won't load slow, etc because of this arp issue (if it is even an issue, as we cannot confirm yet until SWSoft gets back to us).

Tom

[ndndixie]

Thought I'd check early this am to see if the slow traffic time(4am-6am) made any difference:


Ping #1: Got reply in 3.2ms [TTL=59]
Ping #2: Got reply in 2.7ms [TTL=59]
Ping #3: Got reply in 2.6ms [TTL=59]
Ping #4: Got reply in 2.6ms [TTL=59]

[danweber]

Tom,

The mail client probably is behind a firewall but the strange thing is that once it worked subsequent tries will be lightning fast, until such time that no request is being made for an extended time (an hour or so maybe?).

The same with the webpages it self. The pages load quick, once they start loading. Regardless if I try to load the webmail interface or regular content, the first request takes forever but once it starts it's fast.

I have noticed that domains on the same IP only experience the issue with the first request to a specific IP address. If I request domainA it takes a long time until the page starts load. Then I request domainB, which is on the same IP and it loads fast. Then I request domainC, which is on another IP and it loads slow again, until it gets started. This seems to indicate to me that it is an IP related issue, like the IP is "dormant" for a while. Maybe if the VPS is dormant the other VPS on the machine take over resources, causing my VPS to dump most memory to disk and when I start a request it has to be brought back into real memory.

Next I will try a script on an external server that does a request to each of my IPs every 5 minutes and see if I can avoid the idle timeout this way and if that would resolve the issue.

Dan

[charles]

Dan this may be a work around, but certainly not a fix.

We found a dns problem with your VPS. Please try it and let me know if you see an improvement. As far as I can tell, this has solved any delays I was seeing before.

charles

[danweber]

Tom,

The system runs much better now, thanks. I did find the instruction on how to disable ident on smtp/pop and this seemed to have helped along as well. I'll keep an eye out for the performance and see if it stays stable. When I first got the VPS a few months ago it was equally fast but then something must have happened that made it get slow.

For now I am a happy camper again. Your support really rocks and makes a difference!

Dan

[mikelbeck]

I agree, with all of that. My VPS is running fast now, and your support does rock!

Thanks again!

[TomK]

Mike, Dan, Others -

Thanks for the feedback!

As you know in this industry, "Things Happen", they always do and always will in this industry, but I am a firm believer in that what matters most is what a company does and how they (re)act to tackle and solve customer problems.

We are not perfect, no company is, but you have my word and commitment that if/when a problem does arise, we will do our best to resolve it, and if we cannot, we'll be open and honest and tell you that and why, and what your options are, etc.

Customer service is what this company was founded on and is focused on, after all, you guys are our marketing team and have been directly responsible for our growth and success over the years based on your word of mouth referrals - and we thank you for that!

If you have any further problems, please let us know.

[Fred]

No, there is nothing you can do. We are working with swsoft on this but don't have a resolution yet.

charles

i've seen that problem too on my vps... and the explanation about idle site could be true in my case.

Charles, to temporary solve the problem, is having a script ( on my desktop or on another server if available ) connecting to the website every 1 min will do the job ???

UPDATE: arg. i haven't read the last page of this topic before posting a reply. I will ask support to look into it then...

[charles]

Fred, please send a ticket to support. Refer them to this thread and tell us which vps if you have more than one, and also a good test page.

I got a response from SWSoft and they want more info, so the more reproducable test cases i can give them the better.

But to answer your question, i think pinging it would be a bandaid. If you can give us a few days and *not* do this, it will help us track down the real problem and fix it for good.

Anyone else reading this - i need a reproducable test case, and if it only happens certain times of the day (i assume your slow times) I need to know when.

thanks
charles

[mikelbeck]

Anyone else reading this - i need a reproducable test case, and if it only happens certain times of the day (i assume your slow times) I need to know when.

thanks
charles

Here's a very low-use domain that I see this happening on frequently:

www.makeoutsongs.com

[Fred]

yeah, the bandaid can wait since i didn't received any complaints from customers...

Ticket sent to support.

[Fred]

After support response, it looks like my problem isn't related to yours. Sorry for time lost.

My ISP dns is probably the cause. That could explain why none of my customers filled a complain about it.

[charles]

No, this is great news (for you especially)!

Glad your all taken care of.

charles

[mikelbeck]

Been having some speed-related issues today...

Here's an MTR from one of my machines to my VPS:

HOST: hpgmbeck4 Loss% Snt Last Avg Best Wrst StDev
1. 192.168.0.1 0.0% 25 1.1 1.2 1.1 1.8 0.2
2. 10.46.64.1 0.0% 25 9.2 10.0 7.0 15.7 2.1
3. dstswr1-vlan2.rh.stjmny.cv.n 0.0% 25 8.9 9.3 7.6 13.0 1.3
4. r1-ge0-2-0.mhe.hcvlny.cv.net 0.0% 25 10.0 11.4 8.4 35.5 5.4
5. r1-srp0-0.wan.hcvlny.cv.net 0.0% 25 9.2 13.0 8.0 67.8 12.6
6. r2-srp1-0.in.nycmny83.cv.net 0.0% 25 10.6 10.8 9.2 15.7 1.5
7. ??? 100.0 25 0.0 0.0 0.0 0.0 0.0
8. nyk-bb2-pos1-3-0.telia.net 28.0% 25 17.1 17.2 14.9 28.8 3.2
9. ash-bb1-pos7-0-0-0.telia.net 28.0% 25 15.7 16.0 14.9 17.0 0.7
10. so6-1-0.j3.iad.scnet.net 20.0% 25 16.8 18.0 15.2 25.5 2.8
11. ae1-45.j1.iad.scnet.net 28.0% 25 17.9 17.4 15.3 20.4 1.3
12. dvlabs-1.gw.iad.scnet.net 28.0% 25 16.2 47.3 15.2 179.8 53.3
13. unknown11.0.157.204.defender 8.0% 25 16.2 17.3 14.9 27.4 2.6
14. unknown144.14.157.204.defend 8.0% 25 15.8 16.3 14.2 27.2 2.6

You can see there's quite a bit of packet loss there that is outside of my ISP's network.

I'll open a ticket for this as well.

[mikelbeck]

The speed is back to where it should be, but now I see it's taking a different route?

HOST LOSS RCVD SENT BEST AVG WORST
192.168.0.1 0% 16 16 1.07 1.11 1.19
10.46.64.1 0% 16 16 7.33 8.75 12.13
dstswr1-vlan2.rh.stjmny.cv.net 0% 16 16 7.23 8.75 11.53
r1-ge0-2-0.mhe.hcvlny.cv.net 0% 16 16 8.39 41.58 148.55
r1-srp0-0.wan.hcvlny.cv.net 0% 16 16 8.26 10.18 13.56
r2-srp1-0.in.nycmny83.cv.net 0% 16 16 9.11 52.03 206.07
??? 100% 0 16 0.00 0.00 0.00
p13-0.core01.phl01.atlas.cogentco.com 0% 16 16 11.86 76.13 234.53
p4-0.core01.dca01.atlas.cogentco.com 0% 16 16 15.34 48.28 224.23
p2-0.core01.iad01.atlas.cogentco.com 0% 16 16 15.31 17.40 26.09
defender.as.eqxashva.aleron.net 0% 16 16 15.32 57.77 217.55
unknown11.0.157.204.defenderhosting.com 0% 16 16 15.33 17.02 19.51
unknown144.14.157.204.defenderhosting.com 0% 16 16 14.53 15.70
19.16

[Robert]

It appers CV.net and Teila has been having issues all day today. We've been trying to route around it. It should be much better now according to our network manager.

If you're still having problems, please e-mail support with a current traceroute and we'll pass it to our network folks. :)

[mikelbeck]

It appers CV.net and Teila has been having issues all day today. We've been trying to route around it. It should be much better now according to our network manager.

If you're still having problems, please e-mail support with a current traceroute and we'll pass it to our network folks. :)

Already done... I sent the output from the mtr (listed in my response above) asking why the route was different, but I guess you just answered it for me. ;-)

Did nobody notice this problem until I complained about it? It just seems kind of odd that it was fixed within half an hour of my opening a ticket about it.

[danweber]

It seemed to have improved my speeds too. Because I still experienced some slowdowns every so often, and still have the initial delay to idle sites, I have provisioned a dedicated server to see if maybe my expectations are way out there. I then switched nameserver routing to the dedicated server and run the same domains as on the VPS.

My observations:

Static pages load about the same. Would have to use a scientific approach to rule our browser cache, possible ISP caching and so on. But the difference, if at all, is so small that it doesn't justify the effort to do a lenghty comparison.

Dynamic pages (PHP, Perl) seem a little slower on VPS but not really to a point that it would make a difference.

Plesk control panel. About a factor of 10 faster on the dedicated server. Now I have asked about a certificate issue in a separate thread and Charles said that Virtuozzo is filtering traffic to do some things. Could this be the cause since obviously the dedicated server does not have Virtuozzo?

File transfer. I get about 300KB/s on VPS and 220KB/s on the dedicated. Good stuff PowerVPS!!!!


Bottom line: I like the responsiveness of the control panel on the dedicated server but there are numerous reasons that make PowerVPS stay a winner:

- Support is simply outstanding
- Someone else keeps me up to date of important security updates and usually has it already done when I hear about them.
- Room to grow. I can add more disk space or RAM without having to relocate everything to a new box
- Raid 10 disk, more reliable thant the one disk thingy on the dedicated
- A true image backup of the entire VPS, including OS
- Because of the Virtuozzo technology the OS takes almost no disk space and leaves more for web space, and that's with more packages installed than on the dedicated server


On the dedicated server with 512MB I never see any swap space being used while running the same domains as on my VPS. On the VPS I don't know how to just see my stats in "top" but I see a large swap amount. Could it be that my 512MB guaranteed RAM are being swaped when idle and could that cause the initial delay for requests?


Well, PowerVPS is worth every penny! I'll hold on to the dedicated server for a few more days and then I'll cancel it again. I am here to stay but I had to satisfy my curiosity.

[charles]

Thanks for the nice comparison. We are still working on the slow start for idle pages thing, waiting feedback from SWSoft after giving them a few nice test cases.

charles

[charles]

Here's a very low-use domain that I see this happening on frequently:

www.makeoutsongs.com

This looks like DNS lookup delay on the first run. What we found is timing a get of http://www.makeoutsongs.com could take as long as 10 seconds on the first lookup, but was sub-second thereafter. Wait a few hours and the first time would be slow again.

But do this by IP, and there is never a delay. Or do a dig on www.makeoutsongs.com first, then try get http://www.makeoutsongs.com - no delay.

So then the question becomes why does it take so long for that lookup? The cache had obviously expired so it had to be retrieved again, but I am unsure why it takes so long.

I see you have 5 nameservers defined for this domain, and kind of suspect this is part of the problem. I suggest removing ns5,6,7 and see if it helps improve the initial lookup. Or keep ns1 and one of the other (so you have redundancy) but there is no need for 5 nameservers.

Let me know what you find.

charles

[mikelbeck]

dnsstuff.com reports that the speed for the name servers is pretty fast:

Looking up at ns1.ecshosting.net.... Reports 2 A record(s). 4ms.
Looking up at ns2.ecshosting.net.... Reports 2 A record(s). 4ms.
Looking up at ns5.ecshosting.net.... Reports 2 A record(s). 4ms.
Looking up at ns6.ecshosting.net.... Reports 2 A record(s). 43ms.
Looking up at ns7.ecshosting.net.... Reports 2 A record(s). 4ms.

Try this domain, see if you have the same results (10 second delay). It's a domain that only has 2 name servers (ns1 and ns2): www.funkidsmail.com

[falsedawn]

quote:
Did nobody notice this problem until I complained about it? It just seems kind of odd that it was fixed within half an hour of my opening a ticket about it.
/quote

I have been noticing intermittent speed and response issues in my WHM for a day or two. Hopefully this is resolved now. As it is intermittent, I didn't bother submitting a support ticket.

[charles]

dnsstuff.com reports that the speed for the name servers is pretty fast:

Looking up at ns1.ecshosting.net.... Reports 2 A record(s). 4ms.
Looking up at ns2.ecshosting.net.... Reports 2 A record(s). 4ms.
Looking up at ns5.ecshosting.net.... Reports 2 A record(s). 4ms.
Looking up at ns6.ecshosting.net.... Reports 2 A record(s). 43ms.
Looking up at ns7.ecshosting.net.... Reports 2 A record(s). 4ms.


I observed fast responses with dig when testing too (but this was after it had been looked up). But the slow first load is definately DNs and definately there.


Try this domain, see if you have the same results (10 second delay). It's a domain that only has 2 name servers (ns1 and ns2): www.funkidsmail.com

No delay whateover, even after several hour waits. Try it yourself on any *nix box


# time wget www.funkidsmail.com

charles

[falsedawn]

quote:
It appers CV.net and Teila has been having issues all day today. We've been trying to route around it. It should be much better now according to our network manager.
/quote

Just wondering, if you knew about this, why it wasn't posted in the "Operational Status - Network and Servers"?

[mikelbeck]

:
Did nobody notice this problem until I complained about it? It just seems kind of odd that it was fixed within half an hour of my opening a ticket about it.


I have been noticing intermittent speed and response issues in my WHM for a day or two. Hopefully this is resolved now. As it is intermittent, I didn't bother submitting a support ticket.

I meant why had nobody from PowerVPS notice the problem.

[mikelbeck]

I observed fast responses with dig when testing too (but this was after it had been looked up). But the slow first load is definately DNs and definately there.



No delay whateover, even after several hour waits. Try it yourself on any *nix box


# time wget www.funkidsmail.com

charles

I've changed makeoutsongs.com to use only 3 name servers - ns1 & ns2 (hosted on my VPS), and ns5, which is an off-site secondary. Give it a little while for the DNS update to propagate, then check to see if the 10 second delay is still there.

Why would multiple name servers cause this problem? AFAIK, when you try to resolve a name it doesn't go to each name server listed for a domain. So I can't see why having 1 or 8 name servers would introduce any sort of delay.

[mikelbeck]

Bump....

?

[charles]

The additional nameservers *shouldn't* make a difference, but I do know the recursive looks will take longer to make when you have more, since it does a fill recursive query for each.

I see a slight delay when testing now, but never worse than 1.5 seconds. What are you seeing?

charles

[charles]

I meant why had nobody from PowerVPS notice the problem.

We did. Routing changes dont happen by themselves or immediately. We were working with our router guy all morning on that issue.

charles

[Hvu]

I do notice that my domains that have DNS hosted on the VPS server have a slight delay yet domains hosted at DNSmadeEasy have no delay at all.


http://www.robotics1120.net <-- on VPS DNS
http://www.xiphin.com <-- on DNSmadeEasy

[charles]

Yes, the first get of www.robotics1120.net took 0.5 seconds versus 0.2 seconds for the rest. I too saw no delay for xiphin.com.

I am not sure what would cause the delay in the initial dns lookup other than your vps is a general server and not a machine dedicated to dns. I think 0.3 overhead for an initial dns lookup is more than acceptable though.

Mikelbeck, I suggest you use 2 nameservers only.

charles

[mikelbeck]

Mikelbeck, I suggest you use 2 nameservers only.

charles

Ok, I'll try that. I'll change the domain to use ns1 (on VPS) and ns5 (off server).

[mikelbeck]

Ok, I'll try that. I'll change the domain to use ns1 (on VPS) and ns5 (off server).

I'm still seeing an initial delay on this domain. I've changed it from using ns1 (on VPS) and ns5 (OFF VPS) to now use ns1 & ns2 (both on VPS). I'll check it in a few hours to see how it is.

[charles]

I just tried it and still got a hefty delay with the initial dns lookup. You could try one last thing, which is just use ns5 and ns6 (2 dns servers that are not on the vps). But if there is a problem with DNS on your VPS it will probably affect that test anyway. So maybe you should open a ticket with support to have them double check dns on your vps. One thing we have seen help is remove the search line from /etc/resolve.conf, although i am unsure why this is so.

charles

[danweber]

I did some experimenting on my own and I think I might have found something.

A regular DNS client is sending the query to the configured DNS server and lets the DNS server recursively resolve the query. What this means is that if the DNS server does not have the information cached it will query the root domain server, which will point him to another server which is then being queried and might return yet another server that has to be queried. This is all done by the DNS server, without us, the client, knowing.

You can do a dig +trace www.mydomain.com and it will show you how many name servers had to be queried. You will notice that this dig will not have the same delay. That's because the dig will not instruct the DNS server to recursively resolve the name but instruct the DNS server to return the pointer to the next DNS server instead. Dig will then send the same query to the next DNS server and so on, until the final answer. Because there is no load on the local machine the dig is very fast while the configured DNS server probably has hundreds of those recursive lookups to make and thus creates a delay.

For the fun of it I tried to see if I could use my VPS DNS as my primary DNS on my workstation. You can do a simple test with dig again

Use a less frequently used domain to make sure it is not cached yet:


[root@florida ~]# dig www.alhambrabeachmotel.com

; <<>> DiG 9.2.5 <<>> www.alhambrabeachmotel.com
;; global options: printcmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 47915
;; flags: qr rd ra; QUERY: 1, ANSWER: 2, AUTHORITY: 0, ADDITIONAL: 0

;; QUESTION SECTION:
;www.alhambrabeachmotel.com. IN A

;; ANSWER SECTION:
www.alhambrabeachmotel.com. 14400 IN CNAME alhambrabeachmotel.com.
alhambrabeachmotel.com. 14400 IN A 66.225.218.81

;; Query time: 4196 msec
;; SERVER: 4.2.2.3#53(4.2.2.3)
;; WHEN: Thu Jul 14 11:47:04 2005
;; MSG SIZE rcvd: 74

[root@florida ~]#


As you can see the request took 4196ms.

Now the same request but I am using my VPS DNS server.


[root@florida ~]# dig @204.157.15.73 www.alhambrabeachmotel.com

; <<>> DiG 9.2.5 <<>> @204.157.15.73 www.alhambrabeachmotel.com
; (1 server found)
;; global options: printcmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 22157
;; flags: qr rd ra; QUERY: 1, ANSWER: 2, AUTHORITY: 2, ADDITIONAL: 2

;; QUESTION SECTION:
;www.alhambrabeachmotel.com. IN A

;; ANSWER SECTION:
www.alhambrabeachmotel.com. 14400 IN CNAME alhambrabeachmotel.com.
alhambrabeachmotel.com. 14400 IN A 66.225.218.81

;; AUTHORITY SECTION:
alhambrabeachmotel.com. 14400 IN NS ns42.hostforweb.net.
alhambrabeachmotel.com. 14400 IN NS ns41.hostforweb.net.

;; ADDITIONAL SECTION:
ns41.hostforweb.net. 172109 IN A 66.225.219.9
ns42.hostforweb.net. 172109 IN A 69.61.10.11

;; Query time: 103 msec
;; SERVER: 204.157.15.73#53(204.157.15.73)
;; WHEN: Thu Jul 14 11:48:52 2005
;; MSG SIZE rcvd: 158

[root@florida ~]#


This brings me to the conclusion that the delay is simply caused because the configured DNS server is very popular and has soo man requests that it causes a delay. I might be completely wrong on this one but I tried it a few times with various domains and got the same results. This would also mean that some users might not see any delay, if they use a DNS that has a low load factor. I am probably using one of the worst DNS server (4.2.2.3) when it comes to load but my Comcast DNS servers have not been very stable lately.


Dan

[charles]

Dan, i think your analysis is correct in that the dns delay is unavoidable when the domains are lightly used. But I think its less an issue of overloaded DNS servers, but the fact that they fall out of ther cache because of the light use.

The recusive lookups are why I recommended he drop back to 2 dns servers as each one adds time to the lookup (the recursive process happens for all of them), and he did see an improvement.

charles

[mikelbeck]

I just tried it and still got a hefty delay with the initial dns lookup. You could try one last thing, which is just use ns5 and ns6 (2 dns servers that are not on the vps). But if there is a problem with DNS on your VPS it will probably affect that test anyway. So maybe you should open a ticket with support to have them double check dns on your vps. One thing we have seen help is remove the search line from /etc/resolve.conf, although i am unsure why this is so.

charles

ns5 and ns6 are SECONDARY DNSs, would that make any difference if I list them as the only DNSs for the domain?

[charles]

No, as the term secondary is really a misnomer. The dns servers are queried in a round robin fashion, which means if you have 5 DNS servers, each one answers every 5th query.

They are still authoritative which is all that matters.

Truth is I dont expect to see any difference than the ns1/ns2 though.

charles

[mikelbeck]

No, as the term secondary is really a misnomer. The dns servers are queried in a round robin fashion, which means if you have 5 DNS servers, each one answers every 5th query.

They are still authoritative which is all that matters.

Truth is I dont expect to see any difference than the ns1/ns2 though.

charles

Ok, I've changed the domain record and the zone record to use ns5 & ns6 as the name servers for makeoutsongs.com.

[Fred]

No, as the term secondary is really a misnomer. The dns servers are queried in a round robin fashion, which means if you have 5 DNS servers, each one answers every 5th query.

oh! that's interresting! i didn't know that...
One of my customer was on a previous server before being on my powervps... and Google bot is still visiting his website on the another server...
His website isn't the main website ( as http://ip wouldn't show his website... )
I'm wordless with this behaviour... it's only google bot ...
Dns change was made like 7 or 10 days ago...
Is it the google dns cache that don't expire very often ?

[mikelbeck]

Here's something interesting...

makeoutsongs.com has ns5 & ns6 as it's name servers. These are my secondaries, not on the VPS:

Looking up at ns5.ecshosting.net.... Reports 1 A record(s). 3ms.
Looking up at ns6.ecshosting.net.... Reports 1 A record(s). 50ms.


Average of all 2 nameservers: 26ms (plus 191ms overhead).

Score: A+


discusscareers.com has ns1, ns2 (both on VPS), ns5, ns6 and ns7 as it's name servers:


Looking up at ns1.ecshosting.net.... Reports 1 A record(s). 3ms.
Looking up at ns2.ecshosting.net.... Reports 1 A record(s). 3ms.
Looking up at ns5.ecshosting.net.... Reports 1 A record(s). 4ms.
Looking up at ns6.ecshosting.net.... Reports 1 A record(s). 50ms.
Looking up at ns7.ecshosting.net.... Reports 1 A record(s). 4ms.


Average of all 5 nameservers: 12ms (plus 241ms overhead).

Score: A

Took off 2 points since ns1.ecshosting.net allows recursive lookups (if lots of people are using the server, it can slow down).
Took off 2 points since ns2.ecshosting.net allows recursive lookups (if lots of people are using the server, it can slow down).


Note the last 2 lines there... ns1 and ns2 apparantly allow recursive lookups, while the rest do not. Could that have something to do with this issue?

[Hvu]

Recursive basically means to allow other people to use your DNS server as their DNS servers. How can i make this sound more logical. You know when you get DSL/Cable they assign you dns servers to use, Basically the same thing.

[mikelbeck]

Recursive basically means to allow other people to use your DNS server as their DNS servers. How can i make this sound more logical. You know when you get DSL/Cable they assign you dns servers to use, Basically the same thing.

Oh, ok. It wouldn't have any effect on the performance of resolving names?

[mikelbeck]

I just tried it and still got a hefty delay with the initial dns lookup. You could try one last thing, which is just use ns5 and ns6 (2 dns servers that are not on the vps). But if there is a problem with DNS on your VPS it will probably affect that test anyway. So maybe you should open a ticket with support to have them double check dns on your vps. One thing we have seen help is remove the search line from /etc/resolve.conf, although i am unsure why this is so.

charles
Is there still a big delay on www.makeoutsongs.com?

[charles]

Not right now (it was about a second), but I'll try again later to make sure.

charles

[vps-vince]

Is there still a big delay on www.makeoutsongs.com?

I find this site is extremely fast from here in UK.
Wish mine was as fast.

- V

[Hvu]

Makeoutsongs have initial delay but then works fine. Check my personal site

http://xiphin.com/

[mikelbeck]

So what's the solution, or the recommended way to have DNS set up?

makeoutsongs.com uses ns5 & ns6 (off-VPS)
funkidsmail.com uses ns1 & ns2 (ON VPS)
discusscareers.com uses ns5 & ns7 (off-VPS)
elite-computing.net uses all 5 DNS

Can somebody check to see which of these has an initial delay? Maybe that'll give me an idea of which is the best (fastest) way to do this.

Ideally, I'd like to use 3 name servers, at least one on VPS and one off VPS.

[Chris]

makeoutsongs.com uses ns5 & ns6 (off-VPS) Yes, a second or 2
funkidsmail.com uses ns1 & ns2 (ON VPS) No
discusscareers.com uses ns5 & ns7 (off-VPS) Yes a second or 2
elite-computing.net uses all 5 DNS No


That's from here in the US.

[shoopie]

Ping my site today at dnsstuff.com and i got a 33ms respond :(

what happend to my 2ms responds :confused:

Ill try later to see if it will change

[mikelbeck]

I'm seeing the same thing. 32ms, even up to 67ms!

In some cases I'm seeing this:

Ping #1: * [No response]
Ping #2: * [No response]
Ping #3: Got reply in 32ms [TTL=54]
Ping #4: Got reply in 32ms [TTL=54]

[charles]

If you think you have a network problem that we need to address, please submit a support ticket showing

1. Your IP. Go to http://whatsmyip.org/ to find out what it is
2. Output of mtr from your pc to your vps. Get winmtr from http://winmtr.sf.net

Without this information we cannot do anything to help.

Neither times quoted above are bad though.

charles

[shoopie]

just came from dnsstuff.com and got some interesting pings...

deftechgroup.com -
Ping #1: Got reply from 205.177.13.10 in 62ms [TTL=51]
Ping #2: Got reply from 205.177.13.10 in 63ms [TTL=51]
Ping #3: Got reply from 205.177.13.10 in 61ms [TTL=51]
Ping #4: Got reply from 205.177.13.10 in 62ms [TTL=51]


powervps.com -
Ping #1: * [No response]
Ping #2: * [No response]
Ping #3: * [No response]
Ping #4: Got reply from 205.177.13.44 in 61ms [TTL=51]


my site -
Ping #1: Got reply from xxx.xxx.xxx.xxx in 33ms [TTL=54]
Ping #2: Got reply from xxx.xxx.xxx.xxx in 32ms [TTL=54]
Ping #3: Got reply from xxx.xxx.xxx.xxx in 35ms [TTL=54]
Ping #4: Got reply from xxx.xxx.xxx.xxx in 32ms [TTL=54]

[charles]

Whats your point though?

1. Unless you are on the same network as that dnsstuff server (where ever it is, these number dont mean anything for you).

2. You are comparing your site to our websites which are singled homed on ***. Your site is on our BGP mix which is not only more resiliant/reduntant, but also will give you the best route out of multiple providers.

3. A ping is no subsititute for a nice mtr output if you want to see the path taken and where time is spent.

charles

[shoopie]

Whats your point though?

1. Unless you are on the same network as that dnsstuff server (where ever it is, these number dont mean anything for you).

2. You are comparing your site to our websites which are singled homed on ***. Your site is on our BGP mix which is not only more resiliant/reduntant, but also will give you the best route out of multiple providers.

3. A ping is no subsititute for a nice mtr output if you want to see the path taken and where time is spent.

charles

Im saying that they were all at 2ms till now. just wondering what happend.

[charles]

No way to tell unless you used mtr or traceroute before and then now, and compare them.

And again, those dont mean anything to *you*. Has the route you take changed? Has your ping time changed? Has your performance changed? If yes, that might be interesting, but then see above :)

charles

[TomK]

just came from dnsstuff.com and got some interesting pings...

deftechgroup.com -
Ping #1: Got reply from 205.177.13.10 in 62ms [TTL=51]
Ping #2: Got reply from 205.177.13.10 in 63ms [TTL=51]
Ping #3: Got reply from 205.177.13.10 in 61ms [TTL=51]
Ping #4: Got reply from 205.177.13.10 in 62ms [TTL=51]


powervps.com -
Ping #1: * [No response]
Ping #2: * [No response]
Ping #3: * [No response]
Ping #4: Got reply from 205.177.13.44 in 61ms [TTL=51]


my site -
Ping #1: Got reply from xxx.xxx.xxx.xxx in 33ms [TTL=54]
Ping #2: Got reply from xxx.xxx.xxx.xxx in 32ms [TTL=54]
Ping #3: Got reply from xxx.xxx.xxx.xxx in 35ms [TTL=54]
Ping #4: Got reply from xxx.xxx.xxx.xxx in 32ms [TTL=54]

You are trying to troubleshoot a problem, without even knowing if a problem exist.

The site that DNSSTUFF is using to ping looks to be sitting out in California, here is their output to one of our VPS nodes:

Hop T1 T2 T3 Best Graph IP Hostname Dist TTL Ctry Time
1 1 1 2 1.3 ms
206.196.11.254 host254.206.196.11.maximumasp.com. 64 US [Router did not respond]
2 2 2 1 1.5 ms [+0ms]
216.26.129.91 core-rtr02.ge-1-3-0.sdf.xodiax.net. 0 miles [+0] 252 US Unix: 23:25:26.584
3 2 2 * 216.26.129.91 core-rtr02.ge-1-3-0.sdf.xodiax.net. 0 miles [+0] 252 US Unix: 23:25:26.693
4 2 2 2 2.0 ms [+0ms]
216.26.128.229 edge-rtr01.ge-0-2-0.sdf.xodiax.net. 0 miles [+0] 252 US [Router did not respond]
5 57 60 59 50 ms [+48ms]
216.248.0.177 unknown.telcove.net 0 miles [+0] 252 US [Router did not respond]
6 51 52 51 50 ms [+0ms]
24.56.96.33 g-2-1.c1-pitb.telcove.net. 0 miles [+0] 249 US Unix: 23:25:28. 87
7 52 50 50 50 ms [+0ms]
24.56.96.78 p03-4-4.c1-atln.telcove.net. 0 miles [+0] 248 US [Router did not respond]
8 51 51 50 50 ms [+0ms]
24.56.97.30 g-2-1.p0-atln.telcove.net. 0 miles [+0] 249 US [Router did not respond]
9 84 86 85 63 ms [+12ms]
198.32.124.130 ge3-1.cr02.mia02.pccw***.net. 0 miles [+0] 239 US [Router did not respond]
10 82 84 83 63 ms [+0ms]
63.218.44.169 pos4-1.colo02.ash01.pccw***.net. 0 miles [+0] 244 US [Router did not respond]
11 63 67 66 63 ms [+0ms]
63.218.94.90 dvlabs.ge6-10.colo02.ash01.pccw***.net. 0 miles [+0] 245 US [Router did not respond]
12 64 65 64 64 ms [+0ms]
204.157.0.150
[Reached Destination]eqash30.defenderhosting.com. 0 miles [+0] 54 US Unix: 23:24:48.748


Here is a reutrn trace back to them:

[root@vz30 root]# traceroute 206.196.11.254
traceroute to 206.196.11.254 (206.196.11.254), 30 hops max, 38 byte packets
1 colo.aleron.com (204.157.0.1) 0.697 ms 0.421 ms 0.296 ms
2 ge2-10.as.eqxashva.aleron.net (205.198.14.245) 0.512 ms 1.853 ms 1.996 ms
3 dcp-brdr-02.inet.qwest.net (205.171.4.17) 0.272 ms 5.751 ms 0.633 ms
4 205.171.251.33 (205.171.251.33) 0.485 ms 3.000 ms 3.549 ms
5 dca-core-02.inet.qwest.net (205.171.8.221) 3.382 ms 1.652 ms 3.499 ms
6 chi-core-01.inet.qwest.net (205.171.20.169) 31.108 ms 26.574 ms 24.511 ms
7 chi-edge-09.inet.qwest.net (205.171.20.122) 25.246 ms 26.127 ms 24.409 ms
8 65.117.168.138 (65.117.168.138) 37.828 ms 36.909 ms 32.961 ms
9 agg-rtr01.eth4-6.sdf.xodiax.net (216.26.128.226) 66.700 ms 62.744 ms 64.377 ms
10 *
[root@vz30 root]#

So after Chicago, I'm not sure where this host is located, but the delay is hop 8 to 9.

You should submit a ticket with DNSSTUFF and ask them how many nodes they test ping/traceroutes to, and if they use an East or West server to do their testing.

To test my theory, I tried to traceroute from DNSSTUFF to a Cisco IP that sits in San Jose, CA and the time is 12ms, proving that their server is in CA or really close to CA.

1 1 1 0 0.8 ms
206.196.11.254 host254.206.196.11.maximumasp.com. 64 US
2 2 1 1 1.2 ms [+0ms]
216.26.129.83 core-rtr01.ge-1-3-0.sdf.xodiax.net. 0 miles [+0] 252 US Unix: 23:34:42.410
3 2 1 1 1.2 ms [+0ms]
216.26.129.83 core-rtr01.ge-1-3-0.sdf.xodiax.net. 0 miles [+0] 252 US Unix: 23:34:42.510
4 2 1 1 1.3 ms [+0ms]
216.26.128.177 edge-rtr01.ge-0-3-0.sdf.xodiax.net. 0 miles [+0] 252 US
5 14 4 4 4.4 ms [+3ms]
12.118.180.9 unknown.att.net 0 miles [+0] 252 US Unix: 23:34:42.775
6 56 58 57 54 ms [+50ms]
12.123.210.22 tbr1-p012902.cgcil.ip.att.net. 0 miles [+0] 244 US
7 54 56 55 54 ms [+0ms]
12.122.10.6 tbr1-cl1.sffca.ip.att.net. 0 miles [+0] 245 US
8 * * *

[mikelbeck]

Here's an MTR from my home machine to my VPS:

|------------------------------------------------------------------------------------------|
| WinMTR statistics |
| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |
|------------------------------------------------|------|------|------|------|------|------|
| wr850g.motorola.com - 0 | 51 | 51 | 0 | 4 | 110 | 0 |
| 10.30.224.1 - 2 | 51 | 50 | 10 | 26 | 150 | 10 |
| dstswr1-vl2.rh.cormny.cv.net - 0 | 50 | 50 | 10 | 19 | 30 | 20 |
| r1-ge0-1-1.mhe.hcvlny.cv.net - 0 | 50 | 50 | 10 | 34 | 411 | 10 |
| r1-srp0-0.wan.hcvlny.cv.net - 0 | 50 | 50 | 10 | 30 | 220 | 40 |
| r2-srp1-0.wan.prnynj.cv.net - 0 | 50 | 50 | 10 | 22 | 111 | 10 |
| r2-srp1-1.in.nwrknjmd.cv.net - 0 | 50 | 50 | 10 | 24 | 90 | 10 |
| 451be0d6.cst.lightpath.net - 0 | 50 | 50 | 10 | 21 | 40 | 20 |
| 0.ge-2-3-0.cr1.nyc3.us.nlayer.net - 0 | 50 | 50 | 10 | 23 | 110 | 10 |
| 0.so-2-1-0.cr1.iad1.us.nlayer.net - 0 | 50 | 50 | 20 | 26 | 70 | 20 |
| ge1-5.hr2.iad1.us.nlayer.net - 0 | 50 | 50 | 20 | 27 | 100 | 20 |
| dvlabs.ge1-8.hr2.iad1.us.nlayer.net - 0 | 50 | 50 | 10 | 31 | 100 | 20 |
| unknown11.0.157.204.defenderhosting.com - 0 | 50 | 50 | 10 | 27 | 130 | 20 |
|unknown144.14.157.204.defenderhosting.com - 0 | 50 | 50 | 20 | 26 | 50 | 20 |
|________________________________________________| ______|______|______|______|______|______|
WinMTR - 0.8. Copyleft @2000-2002 Vasile Laurentiu Stanimir ( stanimir@cr.nivis.com )

And here's one to one of my domains on my VPS:

|------------------------------------------------------------------------------------------|
| WinMTR statistics |
| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |
|------------------------------------------------|------|------|------|------|------|------|
| wr850g.motorola.com - 0 | 51 | 51 | 0 | 2 | 10 | 0 |
| 10.30.224.1 - 0 | 51 | 51 | 10 | 20 | 50 | 20 |
| dstswr1-vl2.rh.cormny.cv.net - 0 | 51 | 51 | 10 | 17 | 40 | 10 |
| r1-ge0-1-1.mhe.hcvlny.cv.net - 0 | 51 | 51 | 10 | 29 | 191 | 50 |
| r1-srp0-0.wan.hcvlny.cv.net - 0 | 50 | 50 | 10 | 39 | 360 | 10 |
| r2-srp1-0.wan.prnynj.cv.net - 0 | 50 | 50 | 10 | 20 | 40 | 10 |
| r2-srp1-1.in.nwrknjmd.cv.net - 0 | 50 | 50 | 10 | 21 | 40 | 20 |
| 451be0d6.cst.lightpath.net - 0 | 50 | 50 | 10 | 22 | 40 | 20 |
| 0.ge-2-3-0.cr1.nyc3.us.nlayer.net - 0 | 50 | 50 | 10 | 24 | 40 | 20 |
| 0.so-2-1-0.cr1.iad1.us.nlayer.net - 0 | 50 | 50 | 20 | 27 | 50 | 20 |
| ge1-5.hr2.iad1.us.nlayer.net - 0 | 50 | 50 | 10 | 29 | 180 | 50 |
| dvlabs.ge1-8.hr2.iad1.us.nlayer.net - 0 | 50 | 50 | 20 | 42 | 151 | 30 |
| unknown11.0.157.204.defenderhosting.com - 0 | 50 | 50 | 20 | 28 | 60 | 40 |
|unknown144.14.157.204.defenderhosting.com - 0 | 50 | 50 | 20 | 26 | 41 | 30 |
|________________________________________________| ______|______|______|______|______|______|
WinMTR - 0.8. Copyleft @2000-2002 Vasile Laurentiu Stanimir ( stanimir@cr.nivis.com )

[mikelbeck]

You are trying to troubleshoot a problem, without even knowing if a problem exist.

The site that DNSSTUFF is using to ping looks to be sitting out in California, here is their output to one of our VPS nodes:

Prior to today, the ping times from dnsstuff to any of my domains or my VPS were 2ms. Today they're anywhere from 32ms to 67ms.

[TomK]

Prior to today, the ping times from dnsstuff to any of my domains or my VPS were 2ms. Today they're anywhere from 32ms to 67ms.

Correct Mike, that is a DNSSTUFF issue, not a DTG/PowerVPS issue, as their ping server is originating from California this week.

[shoopie]

here is what i have...

|------------------------------------------------------------------------------------------|

| WinMTR statistics |

| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |

|------------------------------------------------|------|------|------|------|------|------|

| No response from host - 100 | 8 | 0 | 0 | 0 | 0 | 0 |

| 68.87.170.177 - 0 | 8 | 8 | 0 | 11 | 16 | 15 |

| 68.87.170.9 - 0 | 8 | 8 | 0 | 13 | 16 | 16 |

| 12.244.72.162 - 0 | 8 | 8 | 15 | 17 | 31 | 31 |

| gbr3-p30.dvmco.ip.att.net - 0 | 8 | 8 | 46 | 56 | 63 | 47 |

| 12.122.10.142 - 0 | 8 | 8 | 47 | 60 | 63 | 63 |

| tbr1-p012402.cgcil.ip.att.net - 0 | 8 | 8 | 47 | 60 | 63 | 63 |

| tbr1-cl1.n54ny.ip.att.net - 0 | 8 | 8 | 47 | 60 | 63 | 47 |

| gar1-p300.nylny.ip.att.net - 0 | 8 | 8 | 46 | 56 | 63 | 47 |

| 12.118.94.46 - 0 | 8 | 8 | 47 | 56 | 63 | 47 |

| so-1-0-0.was10.ip.tiscali.net - 0 | 8 | 8 | 62 | 72 | 94 | 62 |

| dvlabs.cirn.net - 0 | 7 | 7 | 62 | 69 | 110 | 62 |

| eqash21.defenderhosting.com - 0 | 7 | 7 | 62 | 69 | 94 | 63 |

| mysite.com - 0 | 7 | 7 | 62 | 66 | 94 | 62 |

|________________________________________________| ______|______|______|______|______|______|

WinMTR - 0.8. Copyleft @2000-2002 Vasile Laurentiu Stanimir ( stanimir@cr.nivis.com )

[elix]

Next time..warp code tags around it.

[shoopie]

Just came from dnsstuff.com and look waht i have :D

my site -
Ping #1: Got reply from xxx.xxx.xxx.xxx in 4.8ms [TTL=59]
Ping #2: Got reply from xxx.xxx.xxx.xxx in 2.3ms [TTL=59]
Ping #3: Got reply from xxx.xxx.xxx.xxx in 2.4ms [TTL=59]
Ping #4: Got reply from xxx.xxx.xxx.xxx in 2.4ms [TTL=59]

My sites is loading faster now then it was yesterday... now thats what i like ;)


powervps.com -
Ping #1: Got reply from 205.177.13.44 in 2.3ms [TTL=59]
Ping #2: Got reply from 205.177.13.44 in 2.2ms [TTL=59]
Ping #3: Got reply from 205.177.13.44 in 2.2ms [TTL=59]
Ping #4: Got reply from 205.177.13.44 in 2.1ms [TTL=59]

see what i mean?

[mikelbeck]

Yep, I'm seeing the same thing today. 2.5 - 2.8ms.

[charles]

But guys, this just shows that the route from dnsstuff is faster. Whoopdie. Means nothing between you and your vps. mtr that and compare to previous values, and then you'll have something meaningful and interesting.

[shoopie]

But guys, this just shows that the route from dnsstuff is faster. Whoopdie. Means nothing between you and your vps. mtr that and compare to previous values, and then you'll have something meaningful and interesting.

So why is my sites loading faster today then it was yesterday? hmmmm

Now thats the difference i see

[charles]

You have shown us nothing to convince me of anything other than the dnsstuff server you are using today has a better route than the one you used yesterday.

If you want to evaluate and compare the speed your site loads, compare that.

[TomK]

i can't explain it any more clear then I have, and you guys just aren't understanding it.

Check the DNSSTUFF *OUTPUT* on the traceroute, and you'll see the originating server that is doing the ping and traceroutes, yesterday they were originating from California, today they are originating from VA/DC on Hopone.

1 0 0 0 0.4 ms
66.36.240.2 AS0
IANA-RSVD-0 c-vl102-d1.acc.dca2.hopone.net. 255 US Unknown: 82086475
2 0 0 0 0.4 ms [+0ms]
66.36.224.242 AS0
IANA-RSVD-0 ge5-0.core1.dca2.hopone.net. 0 miles [+0] 254 US Unix: 17:44:06.651
3 2 2 2 2.3 ms [+1ms]
66.36.224.18 AS3
MIT-GATEWAYS ge3-0.core1.iad1.hopone.net. 0 miles [+0] 253 US Unknown: 83cec7f2
4 * * * 2.7 ms [+0ms]
[Unknown] [Unknown - Firewall did not respond] 0 miles [+0]
5 * * * 2.7 ms [+0ms]
[Unknown] [Unknown - Firewall did not respond] 0 miles [+0]
6 2 3 2 2.7 ms [+0ms]
204.157.0.100 AS0
IANA-RSVD-0
[Reached Destination]eqash25.defenderhosting.com. -1 miles [+0] 59 US Unix: 17:44:14.301


You act like we are making this up and telling you lies, but the proof has been posted here yesterday, and today.

As Charles said, DNSSTUFF has nothing to do if you are hving performance problems or not. If you are, we need to know that so we can look at it, and the only way to do that is to submit a ticket with urls that are showing problems, outside of the initial delay on first load, we know that is an issue and are working on trying to resolve it.

[shoopie]

You have shown us nothing to convince me of anything other than the dnsstuff server you are using today has a better route than the one you used yesterday.

If you want to evaluate and compare the speed your site loads, compare that.

well obviously you guys did something this morning cause cpanel is asking me to update my contact email in one of my domains.

well squash it because everything is running smooth now... thanks! :D

[Tony]

Um... sidenote: cPanel will ask for a contact e-mail anytime it doesn't have one - it's a feature to allow account password recovery.

=)

[shoopie]

Um... sidenote: cPanel will ask for a contact e-mail anytime it doesn't have one - it's a feature to allow account password recovery.

=)

I know that and i did have one for that domain when i logged in to check server load 3 hours ago. I didnt change anything in cpanel.

The reason why i went to dnsstuff.com to ping my sites is because they were slow to load. Everything is loading fast now... thanks!

Dont mind my english because its my 2nd language... sorry.

[Tony]

We understand that. What we're asking you to do, is ping from a fixed location - i.e. your computer - as opposed to a multi-homed site that gives inaccurate output depending on the time of day.

We'd just like you to do <strike>two</strike> three things:

1) Click start->run type "cmd" (if you're using windows xp/2000) or "command" for other windows versions

In the black box that appears; type "ping siteurl.tld". Paste the output into a ticket - you can paste it here to keep track if you want; but that really doesn't help, so a ticket is preferred - that way, we can track what's going on by both your contact details, and the hn/veid.

2) Download winmtr - it's freely available from http://winmtr.sf.net

Extract it, and open it up. Enter sitename.tld into the host box; then click start - let this run for about 5 minutes, then save as html and attach it to the ticket.

--note: replace "siteurl.tld" with the url of your website, and don't include the quotes.

3) Include your IP! This is available from http://www.whatismyip.com

This allows us to identify what is a possible network issue from what is an ISP issue.

That's all we ask.

We may not be able to solve the issue immediately; but we'll never solve it, if all we have to go on is a forum thread using a random site as the 'test bed' that gives inaccurate output depending on which of it's servers you hit, coupled with anonymous names that we can't track without a headache. So, help us help you. Eh?

[mikelbeck]

I was going to do this when I got home this evening. I posted an mtr last night in this thread, I was going to run another one tonight (and post it) to see what the differences were.

[Tony]

I was going to do this when I got home this evening. I posted an mtr last night in this thread, I was going to run another one tonight (and post it) to see what the differences were.

Please also submit a ticket. :)

http://www.powervps.com/support
support@powervps.com

[ndndixie]

Please also submit a ticket. :)

http://www.powervps.com/support
support@powervps.com
So we're supposed to use support for support? Lol, go figure......the craziest things :p .

[elix]

mikelbeck,

Why does the DNSStuff result matter at all to you? What really matters is the result you get from your computer. I use the same ISP as you, Optimum Online and things seem perfectly fine over nLayer. I get ~ 15-20ms from NY. I never experience delays on my VPS either. However, there are times when Optimum Online's network is clogged in some areas, luckily my area is pretty good for this as there aren't many computer savvy people around. But, in the areas where it's clogged you'll experience slow speeds to pretty much anywhere during peak times. Do a traceroute when it's slow and you'll notice high latency on CV's end.

[mikelbeck]

Here's today's mtr output:


|------------------------------------------------------------------------------------------|
| WinMTR statistics |
| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |
|------------------------------------------------|------|------|------|------|------|------|
| wr850g.motorola.com - 0 | 81 | 81 | 0 | 5 | 160 | 0 |
| 10.30.224.1 - 0 | 81 | 81 | 10 | 19 | 150 | 30 |
| dstswr1-vl2.rh.cormny.cv.net - 0 | 80 | 80 | 10 | 16 | 120 | 10 |
| r1-ge0-1-1.mhe.hcvlny.cv.net - 0 | 80 | 80 | 10 | 22 | 200 | 10 |
| r1-srp0-0.wan.hcvlny.cv.net - 0 | 80 | 80 | 10 | 25 | 210 | 10 |
| r2-srp1-0.wan.prnynj.cv.net - 0 | 80 | 80 | 10 | 20 | 111 | 10 |
| r2-srp1-1.in.nwrknjmd.cv.net - 0 | 80 | 80 | 10 | 22 | 131 | 20 |
| 451be0d6.cst.lightpath.net - 0 | 80 | 80 | 10 | 22 | 110 | 20 |
| 0.ge-2-1-0.cr1.nyc3.us.nlayer.net - 0 | 80 | 80 | 10 | 20 | 70 | 20 |
| 0.so-2-1-0.cr1.iad1.us.nlayer.net - 0 | 80 | 80 | 10 | 24 | 110 | 20 |
| ge1-5.hr2.iad1.us.nlayer.net - 0 | 80 | 80 | 10 | 26 | 240 | 30 |
| dvlabs.ge1-8.hr2.iad1.us.nlayer.net - 0 | 80 | 80 | 10 | 29 | 170 | 20 |
| unknown11.0.157.204.defenderhosting.com - 0 | 80 | 80 | 10 | 27 | 190 | 60 |
|unknown144.14.157.204.defenderhosting.com - 0 | 80 | 80 | 10 | 27 | 120 | 20 |
|________________________________________________| ______|______|______|______|______|______|
WinMTR - 0.8. Copyleft @2000-2002 Vasile Laurentiu Stanimir ( stanimir@cr.nivis.com )

[mikelbeck]

mikelbeck,

Why does the DNSStuff result matter at all to you? What really matters is the result you get from your computer. I use the same ISP as you, Optimum Online and things seem perfectly fine over nLayer. I get ~ 15-20ms from NY. I never experience delays on my VPS either. However, there are times when Optimum Online's network is clogged in some areas, luckily my area is pretty good for this as there aren't many computer savvy people around. But, in the areas where it's clogged you'll experience slow speeds to pretty much anywhere during peak times. Do a traceroute when it's slow and you'll notice high latency on CV's end.
It doesn't really matter all that much to me, I just use it as a gauge to see how things are going. During the day I attach to the Internet from 3 different locations, all on CV's network, but from three different entry points. And the speed to my VPS and sites is different from all locations. dnsstuff is a constant, so if I'm trying to get to my VPS and it's very slow, and dnsstuff tells me that it's slow from there as well, I know something's up and it's not just local network congestion.

[mikelbeck]

Tony -

I'm not sure if we're trying to diagnose a problem in this thread or are all just comparing notes as to how things are going at the moment. If I have a problem - any problem - I open a ticket for it immediately. I also come to this forum and post something about it to see if it's just me or if it's affecting everyone.

Sometimes we (users/clients) take comfort in the fact that we're not the only ones having problems, but everybody else is as well. :D

We understand that. What we're asking you to do, is ping from a fixed location - i.e. your computer - as opposed to a multi-homed site that gives inaccurate output depending on the time of day.

We'd just like you to do <strike>two</strike> three things:

1) Click start->run type "cmd" (if you're using windows xp/2000) or "command" for other windows versions

In the black box that appears; type "ping siteurl.tld". Paste the output into a ticket - you can paste it here to keep track if you want; but that really doesn't help, so a ticket is preferred - that way, we can track what's going on by both your contact details, and the hn/veid.

2) Download winmtr - it's freely available from http://winmtr.sf.net

Extract it, and open it up. Enter sitename.tld into the host box; then click start - let this run for about 5 minutes, then save as html and attach it to the ticket.

--note: replace "siteurl.tld" with the url of your website, and don't include the quotes.

3) Include your IP! This is available from http://www.whatismyip.com

This allows us to identify what is a possible network issue from what is an ISP issue.

That's all we ask.

We may not be able to solve the issue immediately; but we'll never solve it, if all we have to go on is a forum thread using a random site as the 'test bed' that gives inaccurate output depending on which of it's servers you hit, coupled with anonymous names that we can't track without a headache. So, help us help you. Eh?

[Tony]

Here's today's mtr output:

Hi mikelbeck; I don't see a ticket from you on the helpdesk...

Looking at the helpdesk; Veena is pretty busy she isn't going to be looking at the forums as you post this - this situation is the same 24/7. Tech's are not around to watch forums waiting for the possibility that some-one will post something that needs our attention. Our job is to manage the helpdesk - if we at any point have a quiet period, we may venture into the forums -; and so the traceback which we'll eventually get around to doing when we see this post will be skewed. This will NOT identify where the issue is.

Please submit tickets so we can get this traced and logged.

[Tony]

I'm not sure if we're trying to diagnose a problem in this thread

Hopefully not...


or are all just comparing notes as to how things are going at the moment.
<snip>
Sometimes we (users/clients) take comfort in the fact that we're not the only ones having problems, but everybody else is as well. :D

And that, I quite happily agree with; but while you guys are all talking about these problems, we're sweating privately because we want to get it fixed. Without you guys actively submitting tickets over stuff like this, we have no way of identifying what is causing it - aside from the pot luck 'try this, try that' routine... which, to be frank, doesn't work.


<snipped bit>
If I have a problem - any problem - I open a ticket for it immediately. I also come to this forum and post something about it to see if it's just me or if it's affecting everyone.


I agree, and I think that's what forums are for - to develop a close knit community of users who can help each other out -, but there there seems to be this growing grey area over when to submit a ticket, and when to post here (if there's an issue affecting your access to the server, be it load speed, high load, sites timing out, whatever - we want to know!), or indeed, do nothing. We'd rather you guys submitted too many tickets than none at all; or worse, just posted in here then wondered what had happened to us when the thread goes for 4 days without a response from a staff member.

I don't know how else to re-iterate what Charles, Tom, Rob and myself have said - We really really really really would like to get to the bottom of this, but it'll mean you guys submitting a few tickets so we can see the trends.

=)

[ndndixie]

Microsoft Windows XP [Version 5.1.2600]
(C) Copyright 1985-2001 Microsoft Corp.


C:\Documents and Settings\Owner>ping

Pinging with 32 bytes of data:

Reply from : bytes=32 time=199ms TTL=55
Reply from : bytes=32 time=186ms TTL=55
Reply from : bytes=32 time=172ms TTL=55
Reply from bytes=32 time=170ms TTL=55

Ping statistics
Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 170ms, Maximum = 199ms, Average = 181ms
I don't have any problems, well not with the server anyways.......

Adding that I'm on dialup at the moment, outside of Memphis, TN so it's more than acceptable from here.

[mikelbeck]

Ok... I'll open a ticket or two with as much information as I can.

[elix]

DNS Stuff is not a good way to find out how fast your site loads. The problems you are experiencing with delays to your site don't appear to be network issues but, they appear to be issues pertaining to the VPS or the ARP issues.

[mikelbeck]

Hi mikelbeck; I don't see a ticket from you on the helpdesk...

There's one there now! KMY-84173-283.

[mikelbeck]

DNS Stuff is not a good way to find out how fast your site loads. The problems you are experiencing with delays to your site don't appear to be network issues but, they appear to be issues pertaining to the VPS or the ARP issues.
Again, I'm not using it to see how fast my site loads, I'm using it to see if the issue I'm having is local or not.

I use dnsreport.com more than I use dnsstuff.com. That gives better information, I think. It checks to make sure the DNSs are responding properly, the mail server is up, etc.

[elix]

Again, I'm not using it to see how fast my site loads, I'm using it to see if the issue I'm having is local or not.

I use dnsreport.com more than I use dnsstuff.com. That gives better information, I think. It checks to make sure the DNSs are responding properly, the mail server is up, etc.

Then why exactly were you complaing about latency, earlier?

[Tony]

Just so you guys can see how important it is that we get a ticket submitted on stuff like this;

In mikelbeck's case, he submitted one, and veena performed a traceback:


| ge1-5.hr2.iad1.us.nlayer.net - 0 | 34 | 34 | 10 | 27 | 70 | 20 |
| dvlabs.ge1-8.hr2.iad1.us.nlayer.net - 0 | 34 | 34 | 10 | 28 | 60 | 20 |
| unknown11.0.157.204.defenderhosting.com - 68 | 34 | 11 | 20 | 25 | 40 | 20 |
|unknown144.14.157.204.defenderhosting.com - 68 | 34 | 11 | 20 | 28 | 50 | 20 |


As you can see from his mtr above, the last 4 hops show that all the packet loss is at the VPS.

We use asynchronous routing - this means, that from your end, everything looks bad on the final hop; it's only when we're able to perform a traceback that we can see where the issue lies:


Hostname %Loss Rcv Snt Last Best Avg Worst
1. unknown11.0.157.204.defenderhosting.com 0% 2281 2281 0 0 0 3
2. ge2-10.as.eqxashva.aleron.net 2% 2250 2281 0 0 17 275
3. r1-ge10-0-1.in.asbnva16.cv.net 2% 2251 2281 0 0 18 405
4. r2-pos2-2.in.nycmnyzr.cv.net 2% 2252 2281 7 7 17 226
5. r2-srp0-1.wan.hcvlny.cv.net 2% 2250 2281 7 7 17 279
6. r1-srp5-0.mhe.hcvlny.cv.net 2% 2249 2280 8 7 8 201
7. dstswr1-ge3-1.rh.cormny.cv.net 2% 2251 2280 7 7 9 157
8. ubr208-fa1-0.cmts.cormny.cv.net 2% 2251 2280 8 8 12 355
9. ubr208-fa1-0.cmts.cormny.cv.net 97% 74 2280 10 9 13 117
10. ???


As you can see above, there is mild packet loss as soon as we get into cv.net; eventually, we get to hop 9 (way outside our network control)- where the error is -, and see a 97% packet loss; in this case, this is what is causing any networking issue's he's facing when he submitted the ticket.

Once we've got that data, we add it to our tracking page, and over the next week or two, we'll slowly see a trend - i.e. all slow loads have x hop in them, or something similar. In essence, we'll then be able to look at the bigger picture instead of just one or two VEs that we know about.

[elix]

Tony,

The packet loss starts @ Aleron:

2. ge2-10.as.eqxashva.aleron.net

Only one of the CV hops has packet loss but it's high.

[Tony]

I stand corrected. (I'd just woken up when I wrote that reply)

There is packet loss, however, a 31 packet loss out of 2281 packets sent is nothing in comparison to a 2206 packet loss (which would cause major issues for anyone using that network pipe).

[Tony]

Only one of the CV hops has packet loss but it's high.

All the hops after aleron are CV hops and have 2% loss. :p

MTR works by pinging each hop it encounters during the trace - they aren't carried over. =)

[elix]

All the hops after aleron are CV hops and have 2% loss. :p

MTR works by pinging each hop it encounters during the trace - they aren't carried over. =)
Oh so, the traceback was an MTR? I understand that in an MTR things aren't carried over...but it really does look like it was carried over in the second one you listed there (traceback).

[Tony]

..but it really does look like it was carried over in the second one you listed there

I agree with you there!

[charles]

Hi Guys

Could all of you who have seen this slow first load problem, please tell me if

1. Your seeing this on different sites that are on the same IP

or

2. Different sites that are on different IPs.

I'm trying to see if you have 2 sites on one IP and that one site with low traffic shows this problem yet another that does get traffic does not, or if they are always on seperate IPs.

thanks
charles

[mikelbeck]

Here's something else I've noticed.

On a site that is just straight HTML or PHP (no database access or anything), the site comes up almost immediately. For example:

http://www.ecshosting.net
http://www.kgdesigns.org
http://www.yourcreditrepairkit.net

But for a site that accesses a database at startup, it takes longer for the site to come up. But once it's up, it's fast. It looks like it's the inital MySQL connection that takes a while:

http://www.ecshosting.net/cart
http://www.makeoutsongs.com
http://www.discusscareers.com
http://www.northeastbracketracing.com

Obviously it should take longer, since it's opening a connection to a database and retrieving data, but it seems like it's only the inital connection, and I don't see this sort of thing on other sites.

Any ideas? Is there some way to optimize MySQL to reduce this connection time?

[mikelbeck]

Here's a little more data.

Using the Web Page Analyzer at http://www.websiteoptimization.com/services/analyze/

http://www.ecshosting.net (no database connection)
Total HTTP Requests: 29
Total Size: 100560 bytes
Connection Rate Download Time
14.4K 77.94 seconds
28.8K 38.97 seconds
33.6K 33.40 seconds
56K 20.04 seconds
ISDN 128K 6.14 seconds
T1 1.44Mbps 0.53 seconds

http://www.ecshosting.net/cart/ (database connection)
Total HTTP Requests: 27
Total Size: 109418 bytes
Connection Rate Download Time
14.4K 85.40 seconds
28.8K 43.00 seconds
33.6K 36.94 seconds
56K 22.41 seconds
ISDN 128K 7.28 seconds
T1 1.44Mbps 1.18 seconds

[mikelbeck]

Hi Guys

Could all of you who have seen this slow first load problem, please tell me if

1. Your seeing this on different sites that are on the same IP


I'm seeing it on different sites on the same IP.

See my post above... All of the sites I mention (aside from ecshosting.net) are on the same IP.

[charles]

Ok, this is good feeeback. Now if you go to one of the mysql based sites and it loads slow the first time, then subsequent loads are faster, what happens when you immediately visit one of the other mysql based sites? Do they still show an initial delay or are they fast at that point? I am assuming they will be fast.

charles

[mikelbeck]

Ok, this is good feeeback. Now if you go to one of the mysql based sites and it loads slow the first time, then subsequent loads are faster, what happens when you immediately visit one of the other mysql based sites? Do they still show an initial delay or are they fast at that point? I am assuming they will be fast.

charles

No, it doesn't seem to make a difference, there's still an inital delay.

[charles]

Very odd, not what I expected at all. Can anyone else confirm the same?

Still noodling it over and getting feedback from SWSoft.

charles

[KARanden]

I can only say, that I can and will confirm mikelbeck's "reports" regarding this slow/lag opening sites on my VPS.

Tony (support) did disable zlib on my vps, which helps sometimes, but today it has been extremely slow for me again. (Ticket PVPS Support #DHP-50873-592).

I understand and accept that, since I'm in Norway. My VPS will not load as fast as for the US users.
But since I can compare to my old host, shared host located in Texas. I find my VPS slow, as slow as that I'm seriously considering going back to my "old" host and just live with the down time from time to time!

Support here is the best I have experienced, but this initial lag, and especially the slowness in WHM/cPanel is something I'm not sure I can "live with" despite the good support. :(

My only "complain" to support, is that I do find some of them to be a little to quick to blame this slowness to the fact that I'm in Norway!!

[drmagoo]

I can only say, that I can and will confirm mikelbeck's "reports" regarding this slow/lag opening sites on my VPS.

Having a very similar experience as well...Really hoping for a good solution.

But since I can compare to my old host, shared host located in Texas. I find my VPS slow, as slow as that I'm seriously considering going back to my "old" host and just live with the down time from time to time!

Same here. Previous shared hosts with servers located in Texas were much faster, but the downtime was an aggravation that led me here.

Support here is the best I have experienced, but this initial lag, and especially the slowness in WHM/cPanel is something I'm not sure I can "live with" despite the good support.

Again, my experince, attitude is much the same. Perhaps a break through soon?

drmagoo

[falsedawn]

I doubt if the problem can be attributed mainly to the MySQL connection, since (as I think I have already mentioned somewhere else), I have a password protected site, and the time it takes for the login box to pop-up from when I enter the URL and hit "return" is often 2 seconds or more. There can't be any mysql action going on in this period, surely?

And yes, CPanel really is painfully slow here, too.

[charles]

I probably sound like a broken record when I say we are doing our best to get to the bottom of this, but we are. This obviously isn't good for us - our goal is happy customers!

I would like to move you guys with the slow load problem to another server to see if the problem goes away (a completely empty server). We don't have an indication that this has anything to do with server load, but want to see if its a shortage of RAM or swapping.

Please send an email to support with the subject "Slow Load Migrate" and I'll get you migrated tomorrow afternoon/evening. Please ask for the ticket to be assigned to me if there is any confusion.

thanks
charles

[Fred]

please, i would like to know what would be the results of the migration... This is a very interresting thread and i would like to know the happy end :)

[Bogdan]

Yes me too.
I just wanted to contact support regarding this issue, but I sow this topic and decided not to.. :)
I am having similar problems here slow Cpanel/WHM and a initial lag on site loading.
It's not a connection issue as this was also reported to me by companies with very good internet connectivity that are hosted on my VPS.

I am waiting to see if you can find a fix soon.

Bogdan

[charles]

Please send an email to support with the subject "Slow Load Migrate" and I'll get you migrated this afternoon/evening. Please ask for the ticket to be assigned to me if there is any confusion.

[Bogdan]

Well charles ... I do have some important live websites running on my VPS right now so I would prefer not to be part of the experiment :)
But I would like to hear the results of it :p

[charles]

I understand your concerns, but I wouldn't call it an experiment. There is no negative to moving other than a short reboot.

charles

[Hvu]

Migrations are seemless you wont notice them, I'm sure your customers wont either it takes around 5 mins.

[danweber]

Even though I am no longer with PowerVPS I still follow this thread. I had posted a few times when this thread started as I had seen the same symptoms as well. Because a resolution was uncertain I had migrated to a dedicated server but if there is finally a resolution I'll consider moving back here because the support here at PowerVPS simply is the best. Keep posting your experiences so I know when it's "safe" to come back.

[mikelbeck]

Has anybody been moved yet? Is there any difference in speed?

I've just opened a ticket asking to be moved - the ticket number is ACO-27959-234.

[FrancescoV]

It seems i'm having the same problem.
I subscribed a Cpanel power 1 vps at the end of june. The site opened very very quicly. No lag in the "first" http call to my site.
In the last days instead i see this: i type url site ( www.netwargamingitalia.net ), then my browser hang up for some seconds ( ranging from 3 to 6 sec. ) and then the site is rendered by my browser.

[mikelbeck]

My VPS has been moved to a different server.

It seems faster, WHM comes up much faster, sites that had delays seem to be coming up quicker as well.

One thing I've noticed... Before I was moved, my memory usage was up around 60%. After the move it was at 17%, but has been slowly climbing, right now it's at 21%. Swap usage is way down as well, I don't recall exactly what it was before but right now it's 0.999%. Server load is way down as well.

I'm gonna keep an eye on the memory usage to see if there's a corrolation between speed & memory usage.

[elix]

My VPS has been moved to a different server.

It seems faster, WHM comes up much faster, sites that had delays seem to be coming up quicker as well.

One thing I've noticed... Before I was moved, my memory usage was up around 60%. After the move it was at 17%, but has been slowly climbing, right now it's at 21%. Swap usage is way down as well, I don't recall exactly what it was before but right now it's 0.999%. Server load is way down as well.

I'm gonna keep an eye on the memory usage to see if there's a corrolation between speed & memory usage.
You appear to be looking at server status in WHM which is the not the correct way to tell how much memory your VPS is using.

[KARanden]

Once again, I can only agree with mikelbeck, including the memory usage.

My VPS was moved last night (Norwgian time), and it the load speed on my domains was/is at this time very good.
WHM/cPanel loadtime is also very good.

The load speed I now see, I can very well accept and live with.
But it will be interesting to see how it will be when more VPS users are added.

[mikelbeck]

You appear to be looking at server status in WHM which is the not the correct way to tell how much memory your VPS is using.

So where would I find that?

And what does WHM show, the overall server memory usage?

[elix]

So where would I find that?

And what does WHM show, the overall server memory usage?
Yes, WHM will show the overall host node memory usage which should not make much difference to you.

If you goto PowerPanel there will be a number called oomguarpages..because it's 4KB pages just times each number by 4 and divide them 1,000 and then you will get the total usage out of the allowed in MBs.

[danweber]

It's interesting to see this improvement on the new server. After I signed up with PowerVPS I had no problems for a few months and then the degradation in performance started. This would suggest that the servers can't quite handle the amount of VPSs that were loaded.

I still have the theory that the problem is the disk access. Backup tasks will take a toll on disk performance and it seemed that the backup tasks took longer each day (guessing from the later time my backup took place each day). Then swapping memory (which the node I was on did quite a bit) relies on fast disk access and infrequently used domains probably got swapped to disk and needed to be loaded again, which could cause an initial delay.

then of course Virtuozzo is still using a 2.4 kernel and a 2.6 kernel with the improved I/O would probably help a lot here too.

I am sure PowerVPS did not oversell the hardware on paper but in real life a node might not be able to take 100% load to keep good performance. The higher the load the higher the overhead must get.

[Bogdan]

I just want to add that the hardware node that I am on also has a 60% memory usage.
Recently I referred someone to PowerVPS ( I hope I done the right thing :) ) and I also had a chance to look at his VPS and I must say it was running much much faster although it was still a Power 1 just like mine. This is when I started warring.
The hardware note that this person was had a 40% memory usage, so maybe we are heating a hardware limit?
Maybe PowerVPS is putting to many VPS on a single server? :confused: Who knows....but I hope to find out...

One question to those how have been migrated: the new hardware node you are on is based on Opteron processors or on Xeon?

Bogdan

[mikelbeck]

Yes, WHM will show the overall host node memory usage which should not make much difference to you.

If you goto PowerPanel there will be a number called oomguarpages..because it's 4KB pages just times each number by 4 and divide them 1,000 and then you will get the total usage out of the allowed in MBs.
Here's what mine is showing right now:

Current Use: 46,548
Soft Limit: 65,536
Hard Limit: 2,147,483,647

What's the difference between "soft" limit and "hard" limit?

[mikelbeck]

One question to those how have been migrated: the new hardware node you are on is based on Opteron processors or on Xeon?

Bogdan

The one I was moved to shows this:


Processor InformationProcessor #1 Vendor: AuthenticAMD
Processor #1 Name: AMD Opteron(tm) Processor 248
Processor #1 speed: 2189.914 MHz
Processor #1 cache size: 1024 KB

Processor #2 Vendor: AuthenticAMD
Processor #2 Name: AMD Opteron(tm) Processor 248
Processor #2 speed: 2189.914 MHz
Processor #2 cache size: 1024 KB

[elix]

Here's what mine is showing right now:

Current Use: 46,548
Soft Limit: 65,536
Hard Limit: 2,147,483,647

What's the difference between "soft" limit and "hard" limit?
Well if you times it by 4 for each thing it's:

Current Use: ~186MB (Your usage)
Soft Limit: 256MB (Guaranteed)
Hard Limit: 2GB (Burst)

Look at the paranthesis.

hth

[Hvu]

Opterons should make your site run faster ;P It did for me. I was on a backup node in from dec 2004 then moved to my new home dual opterons and it was way faster. Mysql/Apache performance will increase as its has been proven many times AMD kills Intel at webhosting.

My Node

[root@server proc]# more cpuinfo
processor : 0
vendor_id : AuthenticAMD
cpu family : 15
model : 5
model name : AMD Opteron(tm) Processor 250
stepping : 10
cpu MHz : 2392.583
cache size : 1024 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
pse36 clflush mmx fxsr sse sse2 syscall mmxext lm 3dnowext 3dnow
bogomips : 4771.02

processor : 1
vendor_id : AuthenticAMD
cpu family : 15
model : 5
model name : AMD Opteron(tm) Processor 250
stepping : 10
cpu MHz : 2392.583
cache size : 1024 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
pse36 clflush mmx fxsr sse sse2 syscall mmxext lm 3dnowext 3dnow
bogomips : 4784.12



[root@server proc]# more meminfo
total: used: free: shared: buffers: cached:
Mem: 8408244224 8155820032 252424192 0 1316904960 2701824000
Swap: 8389775360 4038742016 4351033344
MemTotal: 8211176 kB
MemFree: 246508 kB
MemShared: 0 kB
Buffers: 1286040 kB
Cached: 1593812 kB
SwapCached: 1044688 kB
Active: 4401356 kB
Inactive: 2729488 kB
HighTotal: 4685248 kB
HighFree: 94944 kB
LowTotal: 3525928 kB
LowFree: 151564 kB
SwapTotal: 8193140 kB
SwapFree: 4249056 kB

[Hvu]

Charles says the kernel 2.6 comes out soon and we should be able to get Dual-Core AMD Opterons and Dual-Core Intel EMT64 ;P They are also testing a new RAID controller too :) PowerVPS always on the edge of technology and keeping customers happy.

[Tony]

Opterons should make your site run faster ;P It did for me. I was on a backup node in from dec 2004 then moved to my new home dual opterons and it was way faster. Mysql/Apache performance will increase as its has been proven many times AMD kills Intel at webhosting.

Unfortunately, a lot of people in the hosting industry have an 'intel thing'... you can show them all the numbers in the world that prove the opterons are faster, but they won't believe it. =)

[elix]

Unfortunately, a lot of people in the hosting industry have an 'intel thing'... you can show them all the numbers in the world that prove the opterons are faster, but they won't believe it. =)
Quite an odd thing...Opterons are faster, cooler, and are more cost efficient...I don't see why everybody isn't using them.

[ndndixie]

I wanna be on the Opteron! I love AMD over Intel ANYDAY!

[charles]

I will say this one more time. If you think you have a performance issue, email support with the subject "Slow Load Migrate" and help solve this, otherwise quit whining!

Yes the opterons out perform the intels, but intels are not bad!

hth
charles

[charles]

My VPS has been moved to a different server.

It seems faster, WHM comes up much faster, sites that had delays seem to be coming up quicker as well.

One thing I've noticed... Before I was moved, my memory usage was up around 60%. After the move it was at 17%, but has been slowly climbing, right now it's at 21%. Swap usage is way down as well, I don't recall exactly what it was before but right now it's 0.999%. Server load is way down as well.

I'm gonna keep an eye on the memory usage to see if there's a corrolation between speed & memory usage.

As I told you in the support ticket, you are looking at the server RAM, not your VPS, but cpu use is your own.

And yes the whole point of this exersize (as stated earlier and in the support ticket) is to prove/disprove that this is a issue with RAM/Swap.

charles

[ndndixie]

My server is great........my connection from here sucks today but not a complaint one on the server. Memory goes up and down, peak periods and all but for the most part, my sites load fast, my customers are happy and so am I.

[Hvu]

Once you go opterons you'll never go back ;P Charles says Windows are on Dual-Core Opterons =[[[[ damn sw-soft needs to hurry up and get 2.6 kernel out so linux can have fun too.

[elix]

My server is great........my connection from here sucks today but not a complaint one on the server. Memory goes up and down, peak periods and all but for the most part, my sites load fast, my customers are happy and so am I.
Which node are ya on?

(do a tracert and then look at the second to last hop :))

VPS Node 30, the one I'm on, is very, very good :-).

[charles]

My server is great........my connection from here sucks today but not a complaint one on the server. Memory goes up and down, peak periods and all but for the most part, my sites load fast, my customers are happy and so am I.

Ok, give that intel a big hug then (even though it's a power sucking pig). ;)

charles

[charles]

Once you go opterons you'll never go back ;P Charles says Windows are on Dual-Core Opterons =[[[[ damn sw-soft needs to hurry up and get 2.6 kernel out so linux can have fun too.

Uh, only one of them. Were having trouble migrating people to it because of umm .... issues ...

charles

[elix]

Uh, only one of them. Were having trouble migrating people to it because of umm .... issues ...

charles
Is this a Dual-Dual Core or is just a Single-Dual Core? :D

[charles]

Dual (so 4 cpus). Pricy as hell right now though.

charles

[elix]

Dual (so 4 cpus). Pricy as hell right now though.

charles
I bet that box runs very fast :D

[Hvu]

Uh, only one of them. Were having trouble migrating people to it because of umm .... issues ...

charles


Yeah Windows and 64bit haha joke. But WinXP64 does have more support for drivers and etc. My new laptop is fully 64bit supported.

I bet that box runs very fast :D

;P run seti and see how many minutes to takes to do a set ;P my old P4 1.4ghz took around 2 days. I'm thinking dual dual-core opterons would take 1 or 2 mins.

[ndndixie]

Ok, give that intel a big hug then (even though it's a power sucking pig). ;)

charles
I always hug my electronics...doesn't everyone? :p

[Hvu]

I always hug my electronics...doesn't everyone? :p

I hugged my laptop when it came. I will hug the routers at PVPS when I do a cage tour. My drool will most likely blow out one of the power supplies, so if the network is down for a bit it was me.

[charles]

/me makes a mental note to not let Henry inside the cages.

[mikelbeck]

Yeah, you guys are being a little strange...

But back to the topic.

My VPS has been very fast since moving to the new server. There's doesn't seem to be any lag in the inital connection, I'm back to being very happy. :D :D :D

[Hvu]

(: glad to hear that Mikelbeck, maybe since its an empty server you get to burst the entire system (:

[ndndixie]

Did Mike just call us strange? Hey Charles? Got an old server around, you know, a REAL slow one? :p

Ode To My Server

Sleek and hard
machine of mine
You send little tiny shivers
up and down my spine.

I love it when you send
those little bytes thru the wires
OH MERCY!
how it lights my fires.

Server O' server,
I'm slave to thee
just the mention of your name
is pure ecstasy! :D (yes, it's been a LONGGGGGGGGGGGGG day!)

[mikelbeck]

I'm seeing slowness every so often today... Every once in a while access to my VPS "goes away" for about 15 seconds, then it comes back and it's fine. Here's an MTR during the "gone away" period:


[root@mbeck4 ~]# mtr --report --report-cycles=50 vps.ecshosting.net
HOST LOSS RCVD SENT BEST AVG WORST
192.168.0.1 0% 50 50 1.02 1.10 1.72
10.46.64.1 0% 50 50 7.33 10.97 21.76
dstswr1-vlan2.rh.stjmny.cv.net 0% 50 50 7.49 10.78 39.81
r1-ge0-2-0.mhe.hcvlny.cv.net 0% 50 50 8.39 13.39 101.41
r1-srp0-0.wan.hcvlny.cv.net 0% 50 50 8.49 20.64 203.14
r2-srp5-0.wan.prnynj.cv.net 0% 50 50 9.54 14.96 41.27
r2-srp1-1.in.nwrknjmd.cv.net 0% 50 50 11.66 16.29 42.81
451be0d6.cst.lightpath.net 0% 50 50 12.54 18.13 53.54
0.ge-3-3-1.cr1.nyc3.us.nlayer.net 0% 50 50 11.73 16.80 38.56
0.so-2-1-0.cr1.iad1.us.nlayer.net 0% 50 50 15.42 20.25 37.34
ge1-5.hr2.iad1.us.nlayer.net 0% 50 50 15.87 23.12 69.07
dvlabs.ge1-8.hr2.iad1.us.nlayer.net 0% 50 50 15.93 33.77 203.26
eqash25.defenderhosting.com 0% 50 50 16.08 19.12 25.12
vps.ecshosting.net 0% 50 50 14.82 18.31 30.30


Also, memory usage (reported in cPanel, the usage for the whole server) has been steadily climbing over the last few days. It was 17% (I think) when I was moved onto this server, today it's 40%.

[elix]

1. Optimum Online has one of the most horrendous networks I've ever seen, those dropouts are very common for me as well but, I know for a fact that it's Optimum Online's fault.

2. I hope you know that Linux is designed to use as much RAM as possible? That's the way Linux works, if you don't like that, move over to Winblowz.

[mikelbeck]

1. Optimum Online has one of the most horrendous networks I've ever seen, those dropouts are very common for me as well but, I know for a fact that it's Optimum Online's fault.

The majority of the big numbers are in cv.net, but there was also a pretty large number in nlayer.net.

But I'll agree with you, OptOnline sucks as far as speed goes.


2. I hope you know that Linux is designed to use as much RAM as possible? That's the way Linux works, if you don't like that, move over to Winblowz.

Thanks for the architecture lesson. ;-)

I'm aware of how Linux works. When I moved to this VPS I commented on how the memory usage was very low, Charles told me to keep an eye on it to see if there was a correlation between the speed of accessing my VPS and the memory usage.

[elix]

Actually OptOnline is nice w/ speed in my area, but I'm pretty lucky as a friend that lives a few miles away from me on OptOnline has a horrendous speed. I still do get dropouts, however.

[charles]

I wish I could somehow get you guys to use our ticket system to help us help you...

Yes we have been adding VPS to the server to bring it to capacity and have been monitoring memory use. It's looking very good right now, so thats definately not it. Were not seeing any other problems with the server either, nor other reports.

I believe this a network issue you are experiencing. If you can keep running mtr and check it when you see this problem, and even screen one onb your vps tracing back, that might help. Otherwise email support when it happens. Even if we can't catch a network problem, we can rule out any other possible problems. If this keeps up, send us your source IP and we can setup the mtr back to you if you need help setting it up.

Posting here doesn't help us catch these. Also not replying to the open support ticket isn't helping me track this. BAH!

charles

[ndndixie]

Reverse psychology Charles.......NO ONE USE SUPPORT YOU MUST POST HERE! type things. :D Don't y'all go bothering them with support tickets now..........

[mikelbeck]

Reverse psychology Charles.......NO ONE USE SUPPORT YOU MUST POST HERE! type things. :D Don't y'all go bothering them with support tickets now..........
I was gonna post here, but saw what you said and decided to open a ticket instead. :D

[ndndixie]

LMAO, see Charles? You just gotta use the same logic I use on my kids. "What? You want a Hersey bar before dinner? SUREEEEEEEEE go ahead, eat 10 while you're at it. " Works everytime, they go grab a carrot stick and look at me like I've lost my mind.

[mikelbeck]

LMAO, see Charles? You just gotta use the same logic I use on my kids. "What? You want a Hersey bar before dinner? SUREEEEEEEEE go ahead, eat 10 while you're at it. " Works everytime, they go grab a carrot stick and look at me like I've lost my mind.

Did somebody say something about potato chips?

[Robert]

Hey Charles... I don't have enough work to do and I'm overpaid!

[mikelbeck]

Hey Charles... I don't have enough work to do and I'm overpaid!
Ya know, there's a new ticket you could be attending to... :p

[charles]

LOL. Too funny. Get back to work dammit, all of you!

[ndndixie]

Hey! This is where I come to escape MY clients! I want a vacation! Hey ****? Do you offer reward points? :D

[elix]

LOL. Too funny. Get back to work dammit, all of you!
I don't have any work :)

[ndndixie]

The dog ate mine.

[mikelbeck]

I'm working with a collegue to get his VPS up and running... I seem to be having the same issues with this one, the inital connection is slow. His is on a different subnet, and I've noticed that there's an extra hop to get to his.

Before you tell me - I've already opened a ticket for this. But I'd like to see if it's just an issue with me or if there's a delay getting to this one for other people.

Take a look. I've got ClientExec set up on both VPSs (his and mine). Here's mine:

http://www.ecshosting.net/ce

Here's his:

http://www.longislandwebhosting.net/ce

Try them both out, see which is faster and post your results here. Thanks.

[elix]

It seems that http://www.longislandwebhosting.net/ce is just overall much slower. Have you installed eAccelerator, etc.?

[mikelbeck]

Yes, eAccelerator is running... I've gone through the VPS and have applied the same tweaks to this one that I have to mine (I think). The CPU usage is low, the only thing that I see different between this one and mine is that the overall memory usage (shown in WHM) is higher. Mine shows 37%, this one shows 55.2%.

[elix]

Regarding the traceroute, you do appear to be right that your VPS has 1 less hop. The extra hop for your friend's VPS is this: corea.ash.dvlabs.com, that appears to be a router...

[charles]

http://www.longislandwebhosting.net/ce was only slightly slower to me on the first load, and faster on the second. If he's still having problems (but I think they need to be more severe than what I am seeing) please have him read this thread then submit a support ticket with the approriate subject and ask to have it escalted to Rob or I.

charles

[elix]

It actually appears to be faster now. Have you turned eAccelerator compression off?

[mikelbeck]

It actually appears to be faster now. Have you turned eAccelerator compression off?
I haven't touched a thing.

[elix]

btw if you need me to tweak some things, just PM me with access, I will glady help make things go faster

[mikelbeck]

btw if you need me to tweak some things, just PM me with access, I will glady help make things go faster
Not that I don't trust you with root access or anything... but...

What would you do to make it faster?

[elix]

Not that I don't trust you with root access or anything... but...

What would you do to make it faster?
I would analyze things and see what is causing this slow down of couse...maybe change around a few httpd and mysql settings for better performance, etc. And I don't think you should be too worried about me doing anything wrong to the VPS, I've had a look at Shahzada's VPS and tweaked a few settings--just ask him.

[mikelbeck]

I would analyze things and see what is causing this slow down of couse...maybe change around a few httpd and mysql settings for better performance, etc. And I don't think you should be too worried about me doing anything wrong to the VPS, I've had a look at Shahzada's VPS and tweaked a few settings--just ask him.
PM is on it's way...

[Zaf]

Yes he has indeed helped me by checking my VPS in the very initial stages. I was scared to modify even the simplest of settings. Infact, I gained a great deal of confidence about settings being set right on the VPS, once he had been through my settings and gave an OK certificate after making a few changes here and there.

Shahzada

[elix]

PM is on it's way...
PM replied.

[falsedawn]

I have just noticed that (I imagine due to the virtuozzo kernel upgrade), that the memory usage on my node, as reported in WHM is around 45%. Normally it is 55-65%.

My sites load pretty much instantly now (no initial lag at all). I will be keeping an eye on the memory usage, and if it creeps back to its previous levels and the lag reappears, I think that's a pretty clear indication of what's causing it.

[Bogdan]

Unfortunately you are probably one of the few lucky ones. It's not related to the upgrade.
Maybe they moved some VPSes or they did a memory upgrade.

My hardware node is still showing 55%-60% and the speed is not that great.....

[TomK]

Unfortunately you are probably one of the few lucky ones. It's not related to the upgrade.
Maybe they moved some VPSes or they did a memory upgrade.

My hardware node is still showing 55%-60% and the speed is not that great.....

55% is still quite low for an entire nodes memory load. What is one of your urls I can test? I can check active swap usage too and see if that is causing any load problems.

[mikelbeck]

Here's one you could check out, if you like:

VZ25 - 38.2% memory usage
http://www.ecshosting.net/ce - 0:00:00.530 seconds
http://www.ecshosting.net/support - 0:00:00.901 seconds

VZ11 - 59.9% memory usage
http://www.longislandwebhosting.net/ce - 0:00:05.497 seconds
http://www.longislandwebhosting.net/support - 0:00:05.638 seconds

I've got an open ticket for this, BTW, it was supposed to be escalated to Rob yesterday.

[Bogdan]

Tom you could test afaceribune[dot]com
The "problem" is the slow Cpanel and WHM. It's been reported by a few customers of mine, some are even from USA and Canada.

It's been already suggested by Charles to move to a different hardware node, but unfortunately I don't want that as I have quite a few customers and resellers and I don't want them to notice the different hardware.
Anyway I think that I will just buy a new VPS to try to see if it makes a performance difference.

[Bogdan]

Here's one you could check out, if you like:

VZ25 - 38.2% memory usage
http://www.ecshosting.net/ce - 0:00:00.530 seconds
http://www.ecshosting.net/support - 0:00:00.901 seconds

VZ11 - 59.9% memory usage
http://www.longislandwebhosting.net/ce - 0:00:05.497 seconds
http://www.longislandwebhosting.net/support - 0:00:05.638 seconds

I've got an open ticket for this, BTW, it was supposed to be escalated to Rob yesterday.

Yeap..I can confirm what you say. Recently I refereed someone to PowerVPS.
His VPS ram usage was about 40%, mine is similar.
His hardware node RAM usage was just 37% and the VPS was moving much faster.

But I suppose that we can't all expect to be hosted on such low RAM usage servers.