Posted on 01/10/2002 6:12:44 PM PST by Bush2000
Sure, but why not cite PC Week, or Information Week, or Yahoo, or CNET, other Ziff-Davis sources, or one of the circulars that come out for Networld/Interop/Comdex? Why not a third-party source from any number of technical news agencies? I am all for technical discussion, but holding up a sales pitch as news isn't news. I would say the same thing about a Sun press release. It's just a press release.
"TPC-A and B benchmarks are into the first phase of their ultimate obsolescence.
It is hoped that the TPC-C benchmark will fill the breach left by TPC-A and TPC-B,
but the initial take up of results has been slower than expected.
The TPC's next benchmark, TPC-D, is in its final review stages
before being officially sanctioned, and is being eagerly awaited by many.
All of this change brings with it uncertainty, and in some cases
people are questioning the worth of TPC testing in general.
Inevitably, one hears statements to the effect that TPC benchmarks are a waste of time, and that they are of no value. "
The current top Sun Processor is the UltraSPARC[tm] III Cu 900-MHz processors.
And if total cost is your goal, you *have* to include Linux. This is how you 'cook the books' to create a great announcement for Bill Gates.
I will now leave this thread to it's regularly scheduled free MS advertising.
An ES7000 with 32 processors and 32gb of memory to achieve these results costs in excess of $950,000. That doesn't count the cost of the Service Contract you have to take with one of these things, because NO ONE ELSE can service them. Tack on about another $80k for the yearly service contract.
A Unix server is still MUCH CHEAPER even if the TPCC's aren't as high. It's HIGHLY UNLIKELY and mid to large size shop is going to need those TPCC's for things like PeopleSoft or SAP. I know, I run a PeopleSoft shop. We looked at moving to PeopleSoft on SQL server, it wasn't as cost effective as what we're doing now, even though the TPCC's were higher. We simply don't need those numbers for what we do.
An ES7000 is a good fit for financial, and high-transactional systems. But for straight Database processing and reporting, Unix is the much better value - STILL.
SQLServer2000 under the Enterprise License agreement I have is $4500 per CPU. On a 32 CPU box like an ES7000 that's $144,000. My existing Oracle license is still cheaper than this. Microsoft thinks they're going to win this battle on pure TPCC counts. They're not. Too many shops have a significant investment in their Unix and back end systems (Oracle, etc..) to just pick up and move to SQL Server. The data conversion isn't the tough part here. The tough part is the process re-engineering, data-flow re-engineering and TRAINING.
Microsoft's margins are the highest of any publically traded technology company on the planet, much higher than Oracle's and much, much higher than Sun's.
From the press relase:
"A look at the top ten price/performance results on the TPC-C Non-Clustered benchmark shows that Windows 2000 and SQL Server 2000 occupy every record on the list."
The "Non-" is a typo. Windows has all of the Clustered results, not all of the Non-Clustered results.
Honestly, I do not think there are any Clustered TPC-C v5 results for UNIX.
The clustered results are kind of irrelevant anyway, since it would be almost impossible to partition a real-world database like the TPC-C database can be partitioned.
Hell, the whole of TPC-C is pretty irrelevant. 2002 marks the tenth anniversary of the benchmark. When TPC-C came out, the typical server was 2 or 4 processors running at 35-40 MHz.
The really interesting information on how irrellevant TPC-C is at http://www.tpc.org/information/about/history.asp
"The TPC-E or Enterprise Benchmark effort was initiated shortly after TPC-C was approved in July, 1992. Proponents of TPC-E argued that while TPC-C was significantly more complex and robust that the earlier TPC-A benchmark, but it still wasn't complex enough to stress very large, enterprise-class systems."
"TPC-E failed to garner enough support for three reasons: 1) the TPC already had one OLTP benchmark (TPC-C), and two benchmarks in this space would only generate industry/user confusion; 2) another benchmark would force vendors to expend precious resources to run another benchmark even more expensive than TPC-C; 3) as an enterprise benchmark, TPC-E was only relevant to a relatively small number of companies competing in that space."
In short, TPC-C a database benchmark meant for small 1-4 processor RISC/UNIX systems with <50 MHz CPUs and a couple of dozen JBOD SCSI 1 disks now claims to be an "enterprise" benchmark, but nobody considered it an enterprise benchmark when it was released a decade ago.
Hmm, what does this really say?
If you go to the Transaction Processing Performance Council's website, go find the Top Ten Non-Clustered TPC-C by Performance Version 5 Results. Look at this table. It is pertinent to this discussion that the page actually shows Microsoft SQL Server 2000 running on the Unisys ES7000 as being number seven in the list with 165,218 transactions. Number one on the list was a Fujitsu PrimePower 2000 showing 455,818 tmpC. So while 165,218 might be a world record for Windows, it's hardly a world record. It is also pertinent to note the Fujitsu ran Solaris 8.
Furthermore, it's kind of a squeeze to suggest Groupe Bull is not a costly proprietary vendor. :P
So, this is a sales pitch. Bully for Microsoft Windows 2000 running on a multiprocessor machine. I don't begrudge their entrance into the market, but you really have to read press releases (not just MS, but anybody's) with a grain of salt.
I wondered about this. Microsoft on Compaq topped the Clustered benchmark, not the Non-Clustered one. You also mentioned that TPM-C is ten years old, which is also pertinent to the discussion because the computing power of the computers has increased dramatically. TPM-C came out in 1991 (1992? about the time the DEC Alpha came out) and some people think it's outdated. Certianly this benchmark will become meaningless at some point, just as statements by Bill Gates that no one will need more than 640KB RAM, or Digital saying 64-bit computing wouldn't catch on.
Does it make you feel better to insult people when you can't answer their questions?
No one should be stupid enough to put their database on .Net. That's what MS is after here, getting corporate databases hooked to .Net.
That's an intelligent response. I state a handful of specific, fact based observations, and you suggest putting my head in the sand will make me feel better?
Of course, another interesting observation about the ES7000 is the fact that both HP and Compaq dumped their OEM agreements with Unisys to offer this system, because there is no market for a 32-way WinTel box (at least not yet).
The ES7000 makes a nice display at a trade show, and is good for high-water mark benchmarks, but until customers trust Windows and SQL Server in the datacenter, it will be a niche machine.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.