Linked by David Adams on Tue 22nd Jun 2010 16:10 UTC, submitted by Jeremy Prince
Oracle and SUN Sometimes, Google's search engine does a better job of telling us about IT vendors than the vendors' own public relations and marketing machines, which are often there mostly to deflect questions rather than answer them. So it is with the next commercial and development iterations of Oracle's Solaris Unix operating system.
Thread beginning with comment 431412
To view parent comment, click here.
To read all comments associated with this story, please click here.
foobar
Member since:
2006-02-07

For instance, Sun has the TPC-C world record right now. But IBM claims they still have the record even today! Because "IBM scores higher TPC-C per core, therefore IBM has the world record". Sure, IBM has higher score per core, but please look at TPC-C and see who has the world record. It is Niagara 1.4GHz machines.


You forgot to mention that IBM used 1 machine, and Oracle used a cluster of 12. Even TPC distinguishes between clustered, and non-clustered. Clustered is just plain silly, and there are only 2 in whole history of TPC-C.

BTW, IBMs TPC-C world record machine, a Unix P595 costed 35 million USD list price. With discount, it costs 17 million USD. That is a ridiculous sum. How can you charge 17 million for ONE AIX server? Sick.


That's wrong. The prices in the disclosures include every piece of hardware and software used in the benchmark. That includes cables, switches, x86 machines to generate the traffic, etc.

TPC-C makes it really easy to look at the prices. There's a nice table in the executive summary. You don't need to look through those huge full disclosure reports that are hundreds of pages long.

I saw a few import things when comparing the 2 submissions.

http://www.tpc.org/results/individual_results/IBM/IBM_595_20080610_...

http://www.tpc.org/results/individual_results/Sun/Sun_T5440_TPC-C_C...

If you look at the table for the p595 submission, most of the money was spent on storage. The list price, before discounts, was $21 million for storage. The list price for the p595 was only $12.6 million. I think a 50% discount is reasonable since they gave a 55% discount on the whole package. So a p595 is more like $6.3 million.

Why did IBM spend so much on storage? It was June 2008 when they ran this test. SSDs weren't an option. 1.5 years later Oracle was able to use some SSDs and reduce their storage costs. TPC-C has always been heavy on storage. Even with the SSD advantage, the Oracle list price for storage was almost $13 million.

The other big cost is the DB software. Oracle RAC is quoted at $7.8 million for a 3 year lease. That's right, like a car, when time is up, you have to pay more if you want to continue using it. And no, I'm not talking about maintenance here. At least the DB2 is $2.3 million and you get to keep using it.

Reply Parent Score: 2

Kebabbert Member since:
2007-07-27

Ok, I was wrong. The IBM P595 server which hold the former TPC-C record, did not cost 17 million USD because the cost included:
A) the server
B) extra hardware such as RAM and discs and cables
C) 128 client computers to generate load
D) third party software

So, if we omit the client computers and omit the third party software, and instead only look at the top configured unix server and necessary RAM and cables - then we dont see a price of 17 million USD. The correct price is instead 16.5 million USD.

So you need ONE unix server for 16.5 million USD to be able to reach that high TPC-C numbers that IBM so proudly boasts. Isnt that sick? Almost 17 million USD for ONE top configured unix server? Imagine the price of one top configured IBM Mainframe which are much more expensive than a lousy Unix machine. And the Mainframe is beaten by one x86 server with eight modern Intel/AMD cpus in terms of processing power. And people thinks monopoly is good for the customers? Pay extremely much for heavily overpriced, slow IBM gear? Are people mad?



Regarding the current TPC-C record. Yes, Oracle has the top spot with a cluster of twelve Niagara SPARC machines. The question is: how can IBM still claim they have the current record because IBM used fewer cores? I dont really get it. Can you explain this to me? Or, is it pure IBM FUD as usual?

IBM was the first company to use FUD and are FUD masters:
http://en.wikipedia.org/wiki/Fear,_uncertainty_and_doubt#Definition
I advice everyone to be very cautios when IBM claims something. Before Microsoft, IBM was always the big Evil company. MS has changed strategy and is nicer now. Not as much FUD as before. But IBM has never stopped lying. They are still FUDing a lot.

For instance, IBM claims that one Mainframe can virtualize 1500 of intel x86 servers. That sounds magnificent. Until you look up the facts. IBM assumes that all x86 servers idle and the Mainframe is loaded 100%! If this is ok, then I can claim that my old laptop can virtualize twenty Mainframes (they must idle). I can software emulate Mainframes on my laptop, that is true. But everyone thinks I am lying if I claim that my laptop can virtualize twenty Mainframes, right? What happens if the Mainframes must do some work? Then my old 1GHz Pentium M laptop would crumble. So it is clearly a lie from IBM. We have all seen my links that proves that a x86 cpu is much much faster than a IBM Mainframe CPU. An IBM Mainframe can not virtualize few x86 servers that do some work, let alone 1500 of them! Lies and FUD.

Likewise, IBM states that the new POWER7 servers can virtualize 800(?) intel x86 servers. But if you look up the facts, it is the same things. The x86 servers in the comparison are very old, and they all idle. And the POWER7 server is fully loaded. This is clearly another IBM lie and FUD. What happens if the x86 servers must do some work?

Master of FUD and lies = IBM. Dont trust IBM.

Reply Parent Score: 1