Resource Allocation

[Dryv]

Hi Forumn,

I have a question. If my goal is to not to overcomit resources at all (I know it takes away a huge value of virtualisation) is this possible. So for example if I have a cluster that has a 100 CPU Cores and 1024GB RAM available to it, can I ensure somehow what is allocated to my VMs (whether its actually used or not by the VM) does not exceed this? So I could have 40 x 2CPU/16GB RAM VMs, and assumingly not be over commited?

Again, I know I'll probably get flamed for this question, but its a question thats been asked to me, and I'm not sure of the best way to go about this.

As usual, thanks for your help, in advance.

Dryv!

[Sateesh_vCloud]

Make sure your design balance below formula.

Total Physical CPU & Memory given to Virtual Machines is < (Always less than)      "TO"    Actual (CPU + RAM) - Overhead for host and guests - Forecast

NOTE:   These are generic calculations and actual values based on your Customer REQUIREMENTS

[jrmunday]

Hi Dryv,

Rightsizing is always a topic for discussion and does require a thorough understanding of both your infrastructure, and workloads. It's often easier said than done, because many customers (ie. the end users) don't actually know the requirements themselves and simply ask for more than they need in a "catch all" approach.

As a general rule of thumb, start small and add resources as required .Benchmark to get a baseline for future comparison and use the native performance charts to help you identify any bottlenecks or over-provisioned guests. Depending on your requirements, resource pools and reservations could help you manage resource allocation and still guarantee the required performance when you need it.

In summary, understand your requirements and cater for them.

Cheers,

Jon

[Dryv]

Hi Both,

Thank you for your reponses. Am I right in saying then there is nothing within VMware that will alert me when I have effectively overprovisioned a clusters resource, even if by just 1 CPU or 1GB RAM?

I have a number of physical servers (10) that currently have 2 Physical CPUs (4 Cores total) and each 8GB RAM. This I guess tells me that inorder to provide the same I need to have a cluster that can offer me 40 CPU Cores and 80GB RAM to host these physical servers. I intend to migrate all these physical servers onto the cluster. In terms of the the right sizing part, I would like assume out, that I will give each machine what it currently has, so one can initially complain their VM is running slower.

What I was looking for was some kind of alert that would tell me that I have now allocated out all my physical resource (even though it may only be getting used partially and could host many more VMs). I guess I simply have to manually track this?