HP hoisted its Superdome-Itanium-Windows combination to the top of the premier OLTP scalability benchmark on Tuesday, about a month after originally gaining the top spot and little more than a week after IBM displaced the combo with a system based on its own AIX/RISC/DB2 stack. Also, Microsoft gained ground on its database rivals in 2002, despite a decline in sales for the market as a whole, according to research published Wednesday by Gartner Dataquest.
TPC-C Benchmark Heats Up, Windows Back on Top
2003-05-21 Benchmarks 17 Comments
Aaah, now we have a truely interesting battle being waged.
It’s not SGI who beat a record of benchmark, just 2 article downs of there?
“SGI Altix 3000 Breaks Speed Record”
“The April 24 benchmark [Superdome+Windows2003] that we had was very schedule driven,” said SQL Server group product manager Tom Rizzo, referring to the goal of putting out a benchmark with the Windows Server 2003 launch. “We did no performance tuning on that benchmark. After we released that benchmark, we decided to go back and do some tuning. The number we have now is more the real number.”
You were asking me something, in one of your posts recently?…. 😉
I hope Oracle is looking to do something else beyond their database business? IBM and MS both have a million-other businesses, and when push comes to shove, they will slive prices to get market share. I am thinking Lary Elison is going to be next Scot McNeally!
Yes, if you see the cost of the server+storage+software the IBM costs much less that HP Superdome.
The fact that it seems it cost more is that TPC-C consider the whole costs, i.e. server and traffic generating clients, but companies normally need to buy only the server part…
The important thing to remember when looking at these numbers is that very, very, very few companies would ever need this much horsepower!
But, that aside the most important number is the cost per transaction. The HP/Microsoft SQL 64bit/Server 2003 combo has the lowest cost/transaction.
This is some pretty impressive stuff for both HP and Microsoft.
zzzzz… Soon as my beowolf cluster of three legged aibo’s is complete, I will rule the TPC-C! zzzz…. lol
And in a year or two, won’t the Sony/IBM Cell system enable me with two or three Playstations and a refrigerator be able to beat them all — put together?!!? 🙂
I was asking why you suddenly ducked out of sight when IBM jumped to the top of the TPC benchmark ladder.
As for the benchmark, I don’t believe the results anyway whether it be a *NIX, *BSD or Windows sitting at the top. These benchmarks are done on the assumption that everything will work out right when in actual fact both you and I know the world of computing is more like murphey’s law.
The Windows TPC result is purely on the fact that MCSE trained people always have and always will be cheaper *on average*, however, if you want to get a well qualified and well rounded one, you *will* pay a premium and you labour cost will be higher that the optimistic numbers generated by Microsoft.
As for IBM, well, that was a simple fact that they *finally* cut the cost of the server, *finally* increased the speed and *finally* did something about the terrible speed DBMS had until recently, hence the reason why all previously benchmarks were using Oracle NOT DBMS.
>>>>The Windows TPC result is purely on the fact that MCSE trained people always have and always will be cheaper *on average*, however, if you want to get a well qualified and well rounded one, you *will* pay a premium and you labour cost will be higher that the optimistic numbers generated by Microsoft.
The “cost” in the TPC-C does not include software cost and staffing cost. From another cnet article:
“IBM’s winning system cost $7.6 million, compared with $6.4 million for the HP Superdome running Windows. However, because IBM’s system uses 32 processors compared to 64 for the Superdome, owners of machines with the IBM design could end up paying lower software licensing fees if the license is based on processor count.”
>>>>>As for IBM, well, that was a simple fact that they *finally* cut the cost of the server, *finally* increased the speed and *finally* did something about the terrible speed DBMS had until recently, hence the reason why all previously benchmarks were using Oracle NOT DBMS.
You hit it right on the mark. This has nothing to do with Microsoft vs. IBM —– this is about Microsoft AND IBM vs. SUN (their hardware is too slow so SUN no longer provide tpc-c benchmark) and Oracle (Oracle losing the relational database market share to IBM and Microsoft).
What IBM will do next — is dropping the BEA java suite from their benchmark platform, and put in their own IBM websphere java suite.
Who cares. Its all a marketing scheme, HP and Microsoft are partners, HP is adopting Microsoft Windows 2003 Server for its Servers, I do not know what this means for HP/UX but I personally dont think many people pay attention to benchmarks. It doesnt point me to one way or another, I use what I use and while yes I find Server 2003 to be a very good system, I dont plan to drop Enterprise 8 anythime soon.
>>>>I do not know what this means for HP/UX but I personally dont think many people pay attention to benchmarks.
“But for all its flaws, interest in TPC-C has re-emerged of late, according to Governor. “I don’t know why it’s come back,” he said. Two major customers–a financial services company in Japan and a retailer in the United States–were pinning their decision to buy IBM Unix servers on the benchmark result, he said.”
I agree with everything you say regarding the ‘validity’ of these benchmarks and how much they matter in the so-called ‘real world’. Also, the fact that a Power4+AIX architecture is performing virtually identically to a Itanium+Windows architecture at about the same price point pretty much proves that they are both competing full-on against each other and pushing the limits of performance as far as modern computer science techniques allow. Period.
BTW, I was late answering you in your post a while back – but I didn’t just ‘duck out of sight’… 🙂
Yes, that was a press-release by SGI allright. They have chosen a specific set of benchmarks and are making claims on their unaudited results.
On the other hand, TPC results are audited, and involve transactions measurements.
So we have different machines, running different benchmarks, some of which are audited and others unaudited.
I would deem them uncomparable.
well nice this costs/transaction. But how reliable are these? because not all calculations are being made by the server, the clients also do calculations/transactions. You getting more points of failure this way. With real benchmarks these calculations are being made completely on server. So i really don’t care really about tpc
LOL, yes I know I can be patronising sometimes, in a humorous way 😉
As for SUN, all I can say is that they need to get their shyte together. For example, they touted UFS has logging by default, wrong, I installed Solaris 9 12/02 x86 and logging wasn’t enabled. I had to manually edit the fstab and put in the mounting options column, “logging”.
I’ve said this a number of times, SUN needs to team up with Veritas and bundle VxFS by default with their operating system just as HP-UX does. As for AIX, IIRC, it used JFS.
The models are all dressed outlandishly, and half of their clothes aren’t meant to wear… it’s all spectacle. The high-fashion machines used in these benchmarks don’t speak to most users.
That being said, I hope IBM comes back and kicks serious ass.