Benchmarking hardware setup

It's been a few weeks since my last posts but I've got a bunch in the pipeline coming up.

Firstly, I've got it together to start using the hardware we got a while back. I'm going to be doing some benchmarking, perf testing and playing with various HA technologies, and of course blogging a bunch about what I discover. The hardware was a lot of fun to put together!

This is kind of a reference post, detailing the hardware setup, so I can link to it in all the posts where I use our hardware.

Photos: front and back (no Facebook login required).

Here's what I'll be using, all mounted in a 42U rack:

  • 2 x Dell PowerEdge 1950 servers, each with
    • 2 x quad-core Xeon E5405 Processor2x6MB Cache, 2.0GHz, 1333MHz FSB
    • 8GB 667MHz RAM
    • 2 x 146GB 15K RPM SCSI drive (for OS etc)
    • 4 3GBps NICs
    • Windows Server 2008 Enterprise SP1 64-bit
    • Multiple SQL Server 2008 Enterprise instances
  • (Upper array) Dell PowerVault MD3000i iSCSI array with 2 dual-port controllers
    • 15 x 1TB 7.2K RPM SATA drives
  • (Middle array) Dell PowerVault MD3000i iSCSI array with 2 dual-port controllers
    • 8 x 300GB 15K RPM SCSI drives
  • (Bottom array) Dell PowerVault MD3000i iSCSI array with 2 dual-port controllers
    • 8 x 1TB 7.2K RPM SATA drives
  • Connected with 1GbE through two PowerConnect 5424 24-port iSCSI optimized switches 

The drives will be reconfigured and I'll specify what I'm using in each post.

The SQL Server instances are vanilla installs, with no special parameters, sp_configure options, or trace flags. If I use any I'll detail them in each post.

Time to get using it!

13 thoughts on “Benchmarking hardware setup

  1. Why do I need a facebook account to view these photos? Can’t you just post the photo’s HERE on SqlSkills.com?
    I gave up Facebook over a year ago, and never looked back. (I couldn’t stand everybody tracking everything I did!)

  2. Now that you moved them from FB to SqlSkills, I can see them. Thank You. Looks impressive, I am jealous!
    How much coin did you drop on all this iron?

  3. Hi,

    is there any document that is really proving the advantages of RAID1 and RAID10 over RAID5? with commits per second in a documented setting? we always have this arguing quarrels with our beloved customers over the optimals settings for SQL.

    thanks,
    Andreas

  4. I’m sure there are, but not off the top of my head. It totally depends on the number of drives in the array, stripe size, partition alignment, I/O pattern which will perform best for any given load.

  5. It is too great to be true for a DBA to setup a immpressive machinese for testing , most of us or my team is using vmware/virtualbox for their R&D .

    Do hope that my company can setup something like this..

  6. Will you let us know how you build your middle and bottom arrays when done? Since each array has two dual-port cards in it, I imagine you’ll be trying to get ‘full-duplex’ out of the backplane, but can you do that with only one half of the storage chassis populated?

    I know back in the day when we did this with SCSI we always filled out a storage chassis and split the backplane by using 4-port cards two ports each for each half of the backplane.

  7. Nothing special I’m afraid – I’ve only got a single NIC in each server dedicated to iSCSI. The two RAID controllers in the chassis have two ports each and can master any of the RAID arrays I define. If I had two NICs in the server I’d be able to isolate traffic through each of them I guess.

  8. Hi Paul

    I was wondering whether you could also run a TPC-C / TPC-E benchmark for the system?
    Also, if you reconfigure the RAID volumes and/or Data+Log file placements, you might want to consider running the benchmark again.

    This is of particular interest to us since we are about to deploy a transaction processing system that co-incidentally uses very (no really, VERY) similar hardware.

    From a High-Availability point of view, we currently have in the lab the two Dell 1950 servers using database mirroring (with automated failover using a 3rd light-weight server running SQL Express exclusively for the Witness role). The servers are configured with 4 NICs each and have two bonded NIC Teams per server(2xNIC’s per TEAM). The one team is a dedicated ‘path’ for the Mirroring traffic as to not have the client connection traffic interfere or be interfered by the mirroring traffic.

  9. I’m not planning to get TPC-C/TPC-E running on the system, but if you have all the necessary executables and a very simple way of getting it to run without me having to do a bunch of config, you could send them to me and I’d run it and publish numbers – win-win for us both.

  10. I believe Quest Software’s Benchmark Factory lets you run TPC-C and TPC-E (plus other) standard workloads without too much effort. Haven’t tried myself but that’s what the glossy brochures say.

Leave a Reply

Your email address will not be published. Required fields are marked *

Other articles

Some thoughts on courage

(This is also the Ponderings – editorial – in today’s SQLskills newsletter.) I want to start out this post by sincerely thanking everyone who emailed

Explore

Imagine feeling confident enough to handle whatever your database throws at you.

With training and consulting from SQLskills, you’ll be able to solve big problems, elevate your team’s capacity, and take control of your data career.