How are CPU weightings determined for each plan?

[artagesw]

I understand that the different plans carry different weightings for CPU usage. What are the relative weightings of each of the plans (Power-0, Power-1, Power-2)?

Also, how do you decide how to cap the number of VPS plans on a single server? I have read elsewhere on this forum that it is simple to upgrade to a bigger plan without moving to a different hardware node. Yet this can't always be true, unless you set a hard limit of 16 VPS plans per server. (16 Power-2 plans would max out an 8GB server, since each plan is guaranteed 512MB.)

Finally, I'd like to see what the "guaranteed" and "burst" values are for CPU utilization in MHz. This would seem to depend on the number and power of the plans on a given server. For example, if the relative CPU weightings are 1, 2 and 4, I'm guessing that a Power-0 would get 1/64th of the CPU, a Power-1 get 1/32nd, and a Power-2 1/16th.

Assuming dual 3 GHz CPUs in a server, 1/64th is approximately 96MHz of processing power for a Power-0 plan.

Am I close?

Sam

[Hvu]

How many plans go on a server depends on the VPS usage. If its low usage sites they might put those lower usage sites on one server and load a few more than than higher usage sites. There isnt really a firm set number of VPS per server since each VPS is different in CPU/Memory/DiskIO usage.

[artagesw]

How many plans go on a server depends on the VPS usage. If its low usage sites they might put those lower usage sites on one server and load a few more than than higher usage sites. There isnt really a firm set number of VPS per server since each VPS is different in CPU/Memory/DiskIO usage.

Perhaps, but if so, then the claim that one can always upgrade to a bigger package without moving to another hardware node would be false. If a single server is loaded up with more than 16 Power-0 plans, and suddenly every one of those plans wants to be upgraded to Power-2, then there is not going to be a way to do that without moving some of those plans to a different server.

[Hvu]

I'm going to let a PVPS rep come answer how they upgrade / move servers.

[TomK]

Perhaps, but if so, then the claim that one can always upgrade to a bigger package without moving to another hardware node would be false. If a single server is loaded up with more than 16 Power-0 plans, and suddenly every one of those plans wants to be upgraded to Power-2, then there is not going to be a way to do that without moving some of those plans to a different server.

Correct if a node is at "capacity". If someone wants to upgrade, we look at a number of items on the node before just upgrading them blindly. If we determine the node cannot handle the upgrade, we will have to migrate them to another node.

We look at many things (historical load, current load, disk IO/IOPS/TPS, memory, etc which we have historical data graphed in our monitoring system) - that help us make these decisions.

Without giving away all the details, each node is targeted for a certain number of "slots" - and a slot equals the plan type, box configuration, etc - not just 10 p1's, 10 p2's, etc. Meaning if a P3 is put on a hw node, that takes a certain number of "slots" allocated to that machine - when the machine hits it max slots, we consider it full. And yes, we leave extra slots for things like bursting, upgrades, etc, since they all happen.

It's a great starting point, and just that. There is no scientific methodology to this "madness" - each customer and VE is different, which makes for a fun day/week, like the last week or so here.

I should be sleeping, it's 3:30AM! Pardon my ramblings above if they don't make sense, just let me know!

Thanks!

[artagesw]

Correct if a node is at "capacity". If someone wants to upgrade, we look at a number of items on the node before just upgrading them blindly. If we determine the node cannot handle the upgrade, we will have to migrate them to another node.

We look at many things (historical load, current load, disk IO/IOPS/TPS, memory, etc which we have historical data graphed in our monitoring system) - that help us make these decisions.

Without giving away all the details, each node is targeted for a certain number of "slots" - and a slot equals the plan type, box configuration, etc - not just 10 p1's, 10 p2's, etc. Meaning if a P3 is put on a hw node, that takes a certain number of "slots" allocated to that machine - when the machine hits it max slots, we consider it full. And yes, we leave extra slots for things like bursting, upgrades, etc, since they all happen.

It's a great starting point, and just that. There is no scientific methodology to this "madness" - each customer and VE is different, which makes for a fun day/week, like the last week or so here.

I should be sleeping, it's 3:30AM! Pardon my ramblings above if they don't make sense, just let me know!

Thanks!
Tom, Thanks for the explanation. The question that still is unanswered in my mind is how much CPU is guaranteed per slot (or plan), and how much CPU bursting capability per slot (or plan). In a smoothly operating VPS, what would you say is the rough equivalent CPU of a Power-0 plan? And what are the relative weightings? Does a Power-1 get twice as much CPU resources as a Power-0, etc.?

[artagesw]

Tom, Thanks for the explanation. The question that still is unanswered in my mind is how much CPU is guaranteed per slot (or plan), and how much CPU bursting capability per slot (or plan). In a smoothly operating VPS, what would you say is the rough equivalent CPU of a Power-0 plan? And what are the relative weightings? Does a Power-1 get twice as much CPU resources as a Power-0, etc.?

No reps at PowerVPS willing to answer this? I'd really like to know how much dedicated CPU utilization I can expect from the various plans...

[elix]

Tom, Thanks for the explanation. The question that still is unanswered in my mind is how much CPU is guaranteed per slot (or plan), and how much CPU bursting capability per slot (or plan). In a smoothly operating VPS, what would you say is the rough equivalent CPU of a Power-0 plan? And what are the relative weightings? Does a Power-1 get twice as much CPU resources as a Power-0, etc.?
Each plan has a certain amount of CPU Units and it's to my understanding that each plan up has double the amount...I'm not sure, however.

No reps at PowerVPS willing to answer this? I'd really like to know how much dedicated CPU utilization I can expect from the various plans...
if you want a timely response you need to e-mail sales@powervps.com

[charles]

No reps at PowerVPS willing to answer this? I'd really like to know how much dedicated CPU utilization I can expect from the various plans...

Please read the sticky at the top of this forum. This is not an official sales/support channel and if you really want a response from us, then email sales.

There is no meaningful number I can give you that will answer this to your satisfaction. We offer a 30 day money back guarantee for this very reason. Try it and see for yourself. When customers want to upgrade we let them try it for a few days to see if it meets their expectations for commiting to it (we can do it on the fly or a short reboot in some cases).

As far as the relative differences are, the power-2 is about 65% more powerful than a power-1 and a power-3 is double a power-2. But the higher plans get more burst as the burst allocated is weighted by the power, so its actually more than that. It's very normal for customers to use more than their guarantee a lot of the time - its the whole point if this technology.

hth
charles

[elix]

charles,

is it true that a Power-1 is comparable to a Celeron while bursting?

thanks

[charles]

Roughly speaking, it should feel like a 2ghz celeron, yes. It's about the closest I can come to describing the power in terms people are used to.