IBM vs HP Servers for VMware

[JoJoGabor]

Hi,

Does anyone have any experience of consolidation ratios for IBM servers against HP? I have seen White papers biased towards IBM showing that they can get twice as many VMs running on a IBM X3650 compared to an HP DL580.

ALso any views on the reliability of both vendors' servers?

[Rodos]

Well you are going to have to compare machines with the same CPUs and RAM, otherwise you are just comparing different configurations.

Check out if the vendors publish and VMmark scores, but again, look for the same configs.I suspect you will find the same config will run quite close between the different vendors. What the different vendors provide is different form factors and ways of doing things. Its like SANs, disks all spin at the same speed and properly configured you will get similar speeds out of competitive SANs, the difference between them is how they do their form factors, software, features and functions, support levels; for a simple metric of IOPS per number of disks you are not going to see massive differences.

Both vendors are reliable.

Considering awarding points if this is of use

[Jasemccarty]

Keep in mind, that even with similar CPU/RAM configurations, the backend of things can still be different.

IBM was touting their memory setup on the x3850's and x3950's when the first came out. And I don't know if HP or Dell had anything like that.

I've seen implementations with IBM, HP, Dell, etc. And I've had success with all of them.

Which do I like? Good question. Which do I use in production? Across my full time, part time, and part time consulting jobs, I've got all 3 manufacturers in production.

And they all do well.

And as far as consolidation ratios... I'm at about 8:1 on my 4/5 year old IBM x440/x445 boxes with 8 single cores, and 32GB of RAM.

Jase McCarty

http://www.jasemccarty.com

Co-Author of VMware ESX Essentials in the Virtual Data Center

(ISBN:1420070274) from Auerbach

[Ken_Cline]

For comparably configured systems from any of the "major" vendors (HP, IBM, Dell, Sun, Siemens, etc.), you should expect to see "performance" parity within about 10-15%. In other words, there's not that much difference in the performance of the top-tier systems. Rather than worrying about the raw performance, look into which vendor has the best service and support offerings in your area. It won't do you any good to have a system that runs three more VMs if you can't get it fixed when it breaks (and it will break...eventually).

From what I've gathered here on the boards, and also personal experience, here are some "benefits" of each:

- HP has the "best" built-in management tools with the Integrated Lights Out module

- IBM has the "best" performance due to more engineering effort in the memory subsystem

- Dell has the "best" price

- Sun has the "best" configurations

- Siemens has the "best" support offerings in Europe

Each of these "bests" is totally subjective and I'm sure there are others here on the boards that will agree/disagree. Also - and this is the most important one - the standings will change with the "next big thing". When HP comes out with their "next" system, they'll have performance bragging rights - until Dell releases their next one, until IBM releases their next one, etc.

So...all of these vendors offer systems that you will have a hard time utilizing the full capability of the box. You're going to hit a ceiling with one component (CPU, RAM, Disk, Network) before you hit the ceiling on the others. That bottleneck will prevent you from maximizing the overall system performance. My advice - don't worry about system performance - worry about vendor performance in your particular area.

Ken Cline

Technical Director, Virtualization

Wells Landers

TVAR Solutions, A Wells Landers Group Company

VMware Communities User Moderator

[JoJoGabor]

Thanks for the information, but I was wondering ifg anyone could backup some of IBM's claims about the performance of their x-series servers in a VDI environment.

For example they have stated they acheived the following results using 4x quad core Intel processors with 64GB RAM, running XP based VDI images with office and an equal simulated load::

HP Proliant 580 47:1

IBM x3850 M2 98:1

I find it hard to believe that simply by using IBM you can run twice as many VDI instances.

[MHAV]

Hi,

I doubt It that IBM has a that much better consolidation ratio than HP Servers. We use HP DL580G5 for our ESX-Servers and they are running really good.

Anyway our suggestion is to put in more than 64GB of RAM no matter what kind of hardware you are using (96 better 128 GB) and look at the 6-Core Prozessors that are available.

Good Luck

[Ken_Cline]

All things being equal except the motherboard, I would be very surprised by those results. Are you sure they were comparing apples to apples? If both systems were running the same CPUs, the same RAM configurations, and the same storage configuration (and the same hypervisor), those numbers just don't make any sense at all.

I could believe that there was a performance difference, but NOT that difference in consolidation density. After all, it is ESX that is going to govern the number of VMs that can run within a given resource allocation - NOT the underlying hardware. The HW will impact how quickly those VMs run, not whether they will start.

So, given two systems that are configured identically (same number & type of CPUs, same amount of RAM, same storage, same network, and same VMs configured identically on each host - all running the same hypervisor with the same settings), you should be able to start the same number of VMs on each host. One host may execute the workloads more quickly...

Ken Cline

Technical Director, Virtualization

Wells Landers

TVAR Solutions, A Wells Landers Group Company

VMware Communities User Moderator

[Dave_Mishchenko]

This is the source document for the ratios provided above - http://www.mjf.ie/u_docs/doc_31_9_event.pdf.

[Jasemccarty]

Hey MHAV...

How many guests do you have on your 580 G5 boxes?

Jase McCarty

http://www.jasemccarty.com

Co-Author of VMware ESX Essentials in the Virtual Data Center

(ISBN:1420070274) from Auerbach

[JoJoGabor]

Does anyone have any other comments on the document? I find it very hard to believe, but it's a published whitepaper. ALthough Tecdem do have strong links with IBM.

[SRENMAN]

I've used all the big 3 over the years as well.

IBM seems to have the lead on the enterprise class x3850 over all the rest including the DL580. Lower power consump, faster mem, better scale up, and usually a great price

I personally think the numbers stated above for the ratio's have to be biased, IBM might be 15-20% faster, but not 50%.

Green Racks has been all the talk over the last few years, have you looked at Performace per watt?

A/C units and Electricity aren't gettin' any cheaper!