It's Sunrise at Oracle. But apparently, someone hit Larry Ellison's snooze alarm. The Oracle chief exec was 18 minutes late for the launch of two preconfigured Sparc/Solaris clusters that are part of a "Sunrise" reanimation of the Sparc platform. Late or not, Ellison was eager about the prospects of the Sparc versions of the …
Solaris 11 Express annoucned 2 weeks ago?
Solaris 11 Express was announced at Open World. It released two weeks ago.
"IBM still has the TPC-C world record"?
Earlier, when Oracle had the TPC-C world record, IBMers claimed that "IBM still has the TPC-C world record, because the IBM cores are faster". That is a clear lie and FUD from IBM.
I bet this new world record of 30M tpmc TPC-C is not a world record, according to IBM and IBMers. "IBM still has the world record".
record belongs to oracle...
> "IBM still has the TPC-C world record, because the IBM cores are faster"
I haven't heard such statement from anyone. Record belongs to Oracle, but that's true that IBM's cores are faster. Oracle uses many many more slower cores, but overall performance of the cluster is higher. In fact Oracle needed to use 9 times more servers to beat IBM's result by a factor of only 3.
".....I haven't heard such statement from anyone...."
Well then you have not been here on this site too long.
I think it is funny how IBMers are backing off from the weird remarks they did earlier. Earlier, IBMers said really strange things, pure FUD if you ask me. I objected vigorously, but they persisted. Now, some time later, they realize the claims they did are really strange. So they are backing off, and deny they said such weird things.
In a couple of years, IBM and IBMers will back off, and deny they did say "one POWER6 core is faster, therefore the POWER6 is fastest cpu". Yes, they did say so! First about Niagara T2. Then they said the same thing about the Intel Nehalem. They are now denying they said so about Intel Nehalem. In some time later, they will deny they said so about Niagara T2, too.
If you read here, you will see many strange claims by IBMers. For instance, how they migrated to POWER7 (even when POWER7 was not for sale) and how fast it is.
There is no business like show business
which CIO would accept such a pile of hardware to run a store with 100.000 items?
I agree that TPC-C benchmarks are totally meaningless and should be retired, just like Sun said since many years back. Sun was not interested in TPC-C benchmarks then. But Oracle and IBM apparatnely thinks the TPC-C benchmarks are important.
TPC-C is the 0 to 60 of the IT world and is more a test of optimisation skills than hardware.
And so the benchmark war continues
First a congrats to oracle on a well done benchmark, they have retaken the Clustered TPC-C benchmark throne !
And it's actually quite a feet to get RAC to scale to 27 nodes, I look forward to seeing what they did.
But it looks like Oracle is up to their usual Software license tricks, It's again only leased software for 3 years with websupport only and you do not actually buy the licenses.
Now the prices are: (Price and Support)
Oracle 47.500 10.450
RAC 23.000 5.060
Partitioning 11.500 2.530
With 1728 processors each trickering a 0.25 license (The copy of the oracle licensing document that Firefox had cached on my HD, actually didn't have an entry for the T3, so I kind of went HMM.. but I found the updated one)
So the real prices would be 35.424.000 and 23.379.840 for 3 years support for a total of 58.803.840 now that is quite a bit more than the 24MUSD that is used when you lease the machines.
Also this time Oracle seems to be able to give some fat discounts. 50% versus the 15% that was used in the last submission.
Machine T5440 T3 T5440 T2+
# machines 27 12
tpmc/machine 1.120.359 637.207
tpmc/core 17.506 19.913
tpmc/Thread 2.188 2.489
Let the battle begin, cause IBM gotta respond to this one :)=
If we talk about pricing. IBM need 77.2 million USD to match this TPC-C record, if POWER7 scaled well enough (which I doubt it does).
Also, the response times from IBM seem to be worse:
"When you look at the averages for New Order you will find 0.353 for Oracle vs. 1.137 for IBM. Despite having a faster core, the response time is actually 3.2 times worse on IBM."
It seems that both IBM and Oracle used SSD drives.
With that said, yes it is impressive that Oracle did this with a cluster. Enormous throughput and decent response times. Myself, I am not really interested in this Oracle record. TPC-C is just for showing off. I think TPC-C should be retired.
If POWER7 scales bad, IBM will not respond to this. If IBM responds to this, POWER7 scales better than people think.
It is a good benchmark but..it's more about the solution than the T3-4
"If we talk about pricing. IBM need 77.2 million USD to match this TPC-C record, if POWER7 scaled well enough (which I doubt it does)."
Well you could just reuse the Oracle setup and replace the T3-4 nodes with POWER servers, no problem.
IMHO the solution of the 3x780 is terrible. IBM will have to bring out DB2 Purescale to beat this excellent Oracle benchmark, on the clustered TPC-C benchmark. But there are also other options. If you look at the rperf numbers for a fully configured POWER 797 and compare it to the nonclustered TPC-C results that have been made for POWER7. Then a fully configured POWER 795 should hit around 29 Million tpmc. (This calculation have also been done by Oracle, you can be sure of that, that is why they have made it a 30M tpmc result). Personally I think the power 795 has to little RAM to hit those numbers. With 16TB RAM on the other hand it might even be able to edge past the 'Super cluster'. Specially if they went a little closer to the Response time limits.
"It seems that both IBM and Oracle used SSD drives."
Well yes and there are big differences. If you compare the two POWER 780 TPC-C benchmarks that have been done, then there is a big difference with regards to IO. The New Oracle clustered benchmark uses Solaris COMSTAR as disk servers. Which is absolutely brilliant.
You basically get much of the Intelligent disk array price without having to pay for one. I do on the other hand doubt that this is something that you'd actually connect to a machine like this in RealLife, but then again the Oracle benchmark one uses 11.040 flash modules mounted on 97 Storage Servers with 388 Nehalem-EP processors and close to 800GB of RAM, where as the IBM clustered 780 benchmark uses 224 direct attached, and the non clustered uses 60 SDD modules. So for the Oracle Benchmark we are kind of like back to the tens of thousands of 'storage' devices, now it's just SDD devices rather than spinning disks like on the SD and p595 non-clustered benchmarks.
This is most likely also the explanation for the good response times, lost of lots of storage devices.
Furthermore the way that Oracle have build up this storage system, then it's very price efficient. But then again the whole benchmark is. And I have no doubt that it'll get used to leverage a Exadata products based upon the T3-4.
The thing we as Oracle customers have to look out for is that for the Exadata products is different from the solution benchmarked here, even though the building blocks are much alike. And the prices are also quite different.
If a single IBM P795 hits anything over 20 million tmpc, then I will be impressed. That is at least as impressive as Oracle's TPC-C record, if not even more impressive. At least to me.
I heard that DB2 used in IBM's top record is not a out-of-the-box commodity DB2 but a special version that is not even sold. However that is only hearsay, so I dont take it too seriously. What can you say about this?
(yes it is a question from my side, not a declarative statement. I make it clear it is only something I heard, that needs to confirmed from other sources. I do not claim that about DB2)
tpmc/core 17.506 19.913
Hmmm, 20,000 tpmc/core and 0,25 per license...
That's 80,000 tpmc per Oracle Enterprise license!!!
Nicely done! IBM POWER did 80,000 tpmcs per license about 3 or 4 years ago.
Oracle is catching up very very fast!!! :)
Quick math to !
Last tpmc Benchmark from IBM
10,366,254 tpmC / 192 cores = 53990 tpmcs per core
Who still has the fastest server? Who is still the master of TPC for NON clustered results?
Anyone can build a database cluster.
Do it then. The only reason no one did it in the past was that it was not easy. Oracle and Sun are making it "easy" with these prebuilt solutions.
The naysayers that kept saying that RAC could not scale are now realizing how wrong they were. IBM will not be able to beat this with any of their own tech. They will have to use Oracle.
Cheaper, faster, and more resilient than IBM. Who could ask for more?
BTW, where is HP in this discussion? Oh yeah, they gave up on the high-end of the market...
Oddly enough in TPC-C land the fastest _oracle_ result on non-clustered technology is still HP's Superdome result from back in 2007!
All IBM's higher results are on DB2 and all Sun's higher results use RAC
Rumours are that HP have better results with oracle that oracle won't let them publish!
"...Rumours are that HP have better results with oracle that oracle won't let them publish!..."
I would not be surprised if that remour would be true back in 2007. But I doubt HP can best the Oracle TPC-C record today. No one can in a forseeable future, I guess.
Why go with a single server when I can get a cluster which has HA included ?