Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

Currently, our customers install our application (services, database, etc.). Some of our customers are asking us to host it for them.

We're looking into hosting providers, but I'm not sure how to specify our requirements, now that everything's gone virtual.

In the past, I would simply say something like:

Quad-core CPU, 16GB RAM, 4x146GB disks configured as RAID10, 2xGbps network cards. Two of them, clustered.

Now, of course, everything's virtual, so CPUs don't really exist, storage is usually on a SAN, and we might not actually need clustering if the virtual fabric provides for live migration. Even cold spares might be enough for some of our customers.

How do I update my hardware-oriented specs to cater for a virtual environment?

Note that I'm not asking about cloud hosting -- our application doesn't scale out enough to make that worthwhile (but it does scale up quite well).

share|improve this question

2 Answers 2

Yes as times are changing everyone is talking about virtual. Its not all that different as before though! We still specify pretty much the same sort of requirement ( i dont know your software or application ) but there are a few rules you might want to follow :

  1. CPU - we would specify how many vCPU we would like to give to the VM. Good thing here if the size of the application grows just increase the vCPU.
  2. RAM - its basically like CPU power as well, you could run your baseline and at times if you need more ram just give the VM more ram ( provided you have enough to give it )
  3. Disk Space and Performance - Yes its true we do have a SAN - it also depends on how the SAN is setup, either for performance, reliability and so forth.....
  4. Network Card - Well its debatable here, no doubt we would have a virtual NIC, some applications might want their own dedicated sort of bandwidth. What you could do here is kill 2 birds with 1 stone! Do NIC teaming - combine 2 NIC cards on the hypervisor side and you would have redundancy as well as load balancing.
  5. Server protection - Hypervisor level, True HA/FT is offered by VMware, HA on XenServer ( sorry not too familiar with the others ) You might want to have at least 2 Hypervisor in order to do vMOTION or XenMotion ( moving a VM from one Hypervisor to the other ) but I would strongly suggest looking into HA/FT solutions by "REAL" HA/FT people! Licenses for HA/FT provided by VMware can sometimes go sky high

Hope this gives you a base understanding!!!

share|improve this answer

CChock's answer is pretty good, but in my experience Hardware Requirements for servers is blown waaay out of proportion. With the monitoring and performance tools that are built into ESX (what I'm most familiar with...), I have done one of two things:

-Agree on some mid-way point (say cut memory and CPU in half), and if the server is performing badly, look at the performance monitoring and raise from there.

or

-Build up the server to their specs, with an agreement that after a month's use you and the customer will look at the performance utilization reports and lower the specs for max efficiency,

I usually strongly push the first option because even in the face of fact (performance charts), customers will throw a fit about taking away spec, even if they aren't using it.

Customers (even Service owners within IT) are still not completely comfortable with VMs. (Many just think they're special; some of them are.) And they still trust the vendor's recommendations; sometimes the requirements are statistics- and testing-based, but more often the company just asked what they were using internally or would recommend.

Many grains of salt are to be taken all 'round.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.