BM Seer Unofficial thoughts from an anonymous Sun employee

HP's tuning on TPC-C

Wednesday Feb 21, 2007

OK I've found some on HP's Itanium TPC-C tuning, now to find IBM's info. HP tired hard to get a good TPC-C but IBM must have done a lot more on TPC-C. ...and this is after IBM did a lot to tune SPECint_rate2000 for Power5+. This covered in http://blogs.sun.com/bmseer/entry/judging_by_the_wrong_things.

But clearly HP did a lot on TPC-C, though it seems like you really need to do a lot at a low level to get good database performance for Itanium2. Also I'm not buying the comment that Itanium2 was beaten by IBM because the CPU was not the bottleneck -- HP did lots to improve CPU performance.

...some questions after reading "Squeezing performance out of Itanium":

  • Do you have to have a PhD in Chip design and Compiler technology to tune your database? :)
  • no improvement going from 400GB to a 600GB SGA... And 2x improvement going from 600GB -> 1000GB. Lots of expensive memory only pays of when you get near 1TB of memory. The latest TPC-C result by HP prices memory at more than 2.2 Million dollars?
  • What about "Out of the Box" performance? -20% without profile feedback optimisation and half the performance without profile and 1TB memory.

[3] Comments
Like this post? del.icio.us | furl | slashdot | technorati | digg
Posted at 11:34AM Feb 21, 2007 in Intel follies
Tags:
Comments:

don't have any problems with vendors tweaking and improving performance as we work with benchmarks. Improving OS and compiler which then benefits any customer -- this is the cool benefit of benchmark competition.

But we need to make sure that benchmarks don't stay around too long. TPC-C is 14 years old. ...and IBM seems to be ahead of the pack by a big margin on TPC-C and not on other benchmarks and now they are not publishing some benchmarks to avoid comparison.

As a reminder Sun 7 years ago quit posting TPC-C since it was too simplistic (and said this publicly in a press release when Sun was the world's fastest!).

For the older set of readers, Dhrystones was a common benchmark years ago and then too many vendors did benchmark special optimizations (optimizations which speed Dhrystones and no real workloads) and rendered its results useless for judging system performance.

Is TPC-C there now?

Posted by BM Seer on February 21, 2007 at 01:19 PM PST #

See the BM Seer posting on 22-Feb.
http://blogs.sun.com/bmseer/entry/ibm_too_tricky_for_good

IBM did lots of optimisations to get high results that go way beyond what a DBA would do.

Posted by BM Seer on February 22, 2007 at 12:21 PM PST #

You may be "not buying the comment that Itanium2 was beaten by IBM because the CPU was not the bottleneck" - but as I said it's likely were hard drives. Just today, new TPCC was published - Oracle on Superdome w/HPUX11i. Sure enough, TPCC speed (4,092,799 tpmC vs 4,033,378) AND the number of hard drives (7056 vs 6768 ) are almost identical - but CPU speed certainly isn't ( SpecInt_2000_rate of 2367 vs ~1500 on 128 Itanium cores vs 64 Power5+).

Posted by Igor on February 27, 2007 at 12:44 PM PST #

Post a Comment:
Comments are closed for this entry.