Mike’s Blog

I measured another system at the office today looking at usage on a 2950 with Energy Smart power supplies, and L series processors (50W each).

Idle usage was 2.1A, and when I pushed the system as hard as I could to light up 4 cores the system went to 2.9A of power.  This is .3A higher than the PE2900 system I am looking at deploying.

There were 2 differences, first, L5410 processors - 2 of them and not just one.  The E5420 used in the 2900 is 80W, the L5410 (and L5420) are 50W each.  That 50W *could* account for ~0.43A of power at 115V.

Second difference - 6 7200 RPM SATA disks vs the 4 15K SAS disks in the 2900.

I’ll continue to see if I can get an even closer match to test against, but I am beginning to wonder if I should drop the whole idea of the 2950 with L series processors just because the cost savings in power do not rack up enough to cover the much higher cost in the server ($450-$700).

EDIT: Updated from .23A to .43A as the difference for 50W CPU at 115V - bad math!

Time for some power measurements!

ESXi was the hypervisor involved in the tests.

System installed, 4 virtuals powered on, but not doing anything: 2.2A @115V

System installed, 4 virtuals being installed hitting the I/O system: 2.4A @115V

System installed, 4 virtuals pushing 100% CPU each, no tuning: 2.6A @115V

While the virtuals were pushing the high CPU load, they were also hitting the disk I/O system as well, though not nearly as hard as 4 concurrent installs occuring.

Part 3.2 will continue with real life power measurements of the systems these PE2900s would replace so that I can do a comparison based on idle vs full load against current production systems (mix of idle and load).

Part 3.3 will have information as I cut over a couple of the clustered systems onto the PE2900 virtualized servers starting with one of the web servers and a POP/IMAP server, continuing from there to a couple of the SMTP servers.  I’ll be able to report back subjective performance reactions as well as some actual measured data via different utilities.

While this is happening, I am working on learning how to make my own ‘appliances’ for faster configuration and turn-up of the different servers I’d like to deploy.  It has been kind of boring so far, but maybe I’ll get’er all figured out.

Until next time…

Oh noes, part 3 is upon me and now I must be witty or informative.  I am tired today, so I’ll try for informative.

My PE2900 showed up today and has been put into the rack.  I even have a silly picture of its guts…

Wasn’t that nice?  Say ‘yes’ please, it helps my ego a lot.

Now onto the virtualization ghetto tour!

Read the rest of this entry »

Ah, the dreaded part 2 of the series.

I ordered my test system from Dell on Tuesday, September 30th, 2008, and the box shipped on Thursday, October 2nd, 2008. I am excitedly waiting for it.

The box I ordered is to test out the I/O performance for the mail server virtualization, the config is:

• Quad Core Xeon E5420, 2.5Ghz, 2×6MB L2 Cache, 1333Mhz FSB
• 16GB (8×2GB FBDIMM)
• Quantity 4 146GB, 15K RPM SAS disks
• PERC6/i RAID controller
• Dual Port Intel 1000PT

So, not a huge machine, though not small either. If this works out well, then I’ll order up 8 of them to start the compression of my physical servers into virtual servers.

Read the rest of this entry »

With things changing all over the marketplace, virtualization has, again, come to the forefront as the savior of the data center.

And wouldn’t you know, I’d like to save my data center, at least some power and cooling needs.

I have started to review how we use our servers and where we could do combining to save power, cooling, and rack space.  During this installment, I’ll be discussing the usage, and combining, of 3 parts of our network:

• POP/IMAP servers
• Apache based web servers
• SMTP (inbound, delivery, outbound) servers

Read the rest of this entry »