Me too!

Quicksearch

Disclaimer

The individual owning this blog works for Oracle in Germany. The opinions expressed here are his own, are not necessarily reviewed in advance by anyone but the individual author, and neither Oracle nor any other party necessarily agrees with them.

Navigation

< Autumn | Different CPU, same challenges >

Monday, October 5. 2009

Looks like IBM is somewhat concerned about the Exabyte V2 announcement and that announcement that is like to happen on October 14th. At least there is a loud "We have it, too - kind of" from Armonk. They tout their product IBM DB2 Pure Scale. Power machines with AIX, some infiniband gearand special edition of DB2.

Some educated guesses

Mr. BlueToTheBone is right. Details are sparse at the moment. But just to write down some points, you remember for having a deeper look at after an official announcement. Those points are purely speculative given that the information in the ElReg article are really sparse:

It will be interesting how this solution with a centralized lock management will compare with the distributed lock management in RAC. Perhaps this will be the scaleability stopper for the IBM solution. I think we will see a solution with fewer, but bigger nodes. Given that the job of lock management is consolidated on one node, i would say that the front end node will be a really big machine.
But have the same opinion like Chris Mellor (hell freezes over, i have the same opinion): IBM has a storage problem. There was the 1 Million IOPS Quicksilver marketing stunt. But that system were just 40 Fusion-IO cards in a lot of servers. So that doesn't look like something playing in the same ballpark like the Sun F5100
The QuickSilver flash was composed of one or two 250GB Fusion-io ioDrives with PCIe interfaces in 29 host servers connected by Fibre Channel to the 13 SVCs. There were 41 ioDrives in total.
Just for the uninitated: The SVC (Storage SAN Volume Controller) is a server as well. A full rack of hardware for 1 Million IOPS. But perhaps they just plug the Fusion-IO in the database servers and don't use SVC at all... who knows
The p6 pSeries systems doesn't have PCI Express 2.0. With the PCIe 1.0 variant you have 240 MBytes/s. Thus you have somewhat around 2 GByte/s on the x8 PCIe of a p520 for example. That isn't enough for QDR. Some systems like the 520 provide GX busses for better connectivity, but nevertheless there doesn't seem to be QDR cards supported in the systems. But those 520 are 4 RU systems, don't know if they use them for such a cluster. Thus i would expect to see DDR infiniband in those systems. I assume they will have half the bandwidth in the fabric
There isn't a system in the pSeries lineup that allows to put 12 disks on 2 rack units. That's the other side of the storage problem, albeit Chis Mellor at El Reg doesn't mention it. That's the point why the X4275 is such an interesting system. You can have 12 Disks and 2 sockets on 2 rack units. I assume the database servers have seperate storage arrays.

There are some other points, but those are much deeper in the realm of speculation, so i just wait for the announcement for further comments.

Frankenbase

Some people say, that Exabyte V2 in its current incarnation is just rushed together in the light of the upcoming merger. But the stuff that IBM told exclusively in the ears of Bluetothebone sounds like kludged together just to counter an competitors announcement. I'm sure that IBM will show of a strange configuration, calling it DB2 Pure Scale. But Frankenstein creation will look as a beauty beside this solution.

Posted by Joerg Moellenkamp in English, Technology, The IT Business at 15:19 | Comments (9) | Trackbacks (0)

View as PDF: This entry | This month | Full blog

Trackbacks

Trackback specific URI for this entry

No Trackbacks

Comments

Display comments as (Linear | Threaded)

and who owns the trademark Power 8 (P8)

#1 frank (Homepage) on 2009-10-05 19:31 (Reply)

One thing you have to look out for is that IBM has 32 way Nehalem-EX boxes in development, that is a 4x 8 Socket design. Each 8 Socket Nehalem-EX has 18x PCI-E 2.0 slots, if filled with FusionIO, would be quite a match to Exadata v2.

When is Sun going to dump Sparc, and go 8 way to 32way Nehalem-EX?

#2 TS on 2009-10-05 19:33 (Reply)

Sorry, but:
- Nehalem-EX to that scale has to proove it's scaleability, nobody has really benchmarked this boxes. Especially for "non-embarrasingly parallel tasks". So i find it interesting, that you trumpet for this chip at this volume.
- SPARC will not be dumped. Period.
- As far as i know the memory controller in the Beckton hasn't changes, thus it should downclock when you use more than one dimm per controller.
- Nobody knows how Beckton behaves on real-world load outside of the 24 MByte of cache.
- Even with MCA Nehalem isn't near the capabilities of real RISC CPUs.
- Nothing prevents IBM from taking Nehalems for this task, but article at the Register talked about AIX .... so pSeries ...
- Another interesting question: How does it scale I/O-wise? How are the memory access characteristics and so on....

I will just wait for the first real systems Given the facts that we have Nehalem systems now, i'm pretty sure we have competitive Nehalem-EX systems in the future

#2.1 Jörg+M. (Homepage) on 2009-10-05 20:02 (Reply)

IBM is annoying by 'bluewashing' technology that is not theirs e.g. IBM's 12X is infiniband.

Big boxes are irrelevant : if one goes down and your whole company is on it, it will take the everything on it to lala-land. If you consolidate multicompany they are a DOG to get everybody off the machine in the same weekend of a needed POR or an upgrade.

Now for some good reading :
(I found this gentleman's blog via the realworldtech.com forum)

Why would IBM need such an announcement even if the product will be bogus :
http://perilsofparallel.blogspot.com/2009/02/larrabee-in-ps4-so-whats-future-of-ibms.html
(if in a hurry ff 2 the 'Back in the late 90s, ', about halfway down)

Joerg : a propos agreeing with Chris Mellor : You are in the best of company! :
http://perilsofparallel.blogspot.com/2008/09/io-io-we-really-need-io.html
(which links to : http://www.channelregister.co.uk/2008/09/22/flash_a_cache_or_hdd/)

I probably will get his book(s) after finishing my OpenSolaris Bible : http://perilsofparallel.blogspot.com/2008/09/background-me-and-this-blog.html

* POR : Power On Reset.

#3 Michiel (Homepage) on 2009-10-06 10:05 (Reply)

To clarify (hopefully)
"http://perilsofparallel.blogspot.com/2009/02/larrabee-in-ps4-so-whats-future-of-ibms.html"
I refer to this wonderful anecdote to remind that it's all about the credibility game which ultimately means politics. Never mind the reality (or facts, see infra).

From : http://blogs.zdnet.com/Murphy/?p=1615
"Remember, when thinking about this, that there are two main reasons your bosses golf with these guys:

1. first, there’s social re-assurance: hanging out with these guys confirms to your bosses that they’re entitled to be your bosses and can understand the business of IT without having to deal with science, numbers, or facts; and, "

#3.1 Michiel on 2009-10-06 10:47 (Reply)

Woes to IBM; guess they should just pack up and quit the business now eh?

The SUNACLE box looks impressive despite Ellison's presentation. V2 may very well trounce IBM kits for some time, but for how long? ( déjà vu AMD & Intel )

Keeping a substantial lead for 3 years is a whole different ball game and set of bragging rights.

disclaimer: While I prefer Oracle to DB2, Sun's hardware has a fairly consistent history of overpriced 'paltry' performers.

#4 vizzini on 2009-10-06 17:50 (Reply)

Is DB2 Purescale just a reaction to the announcement from Oracle? Of course it is. IBM probably don't expect to sell anymore of those than was sold of Exadata1. IBM will just make sure that Oracle won't hold the top spot of the TPC-C clustered benchmark for very long.

Up till now IBM and Oracle has had a nearly symbiotic relationship. IBM sells expensive servers, Oracle sells even more expensive software that runs on them. Despite Larry Ellisons obviuos IBM envy there is a huge collaboration between IBM and Oracle to make the Oracle DB run optimal on Power gear. Most Oracle techies I've spoken to attest to the virtues of this combination and they may also tell you a few things off the record about RAC that you probably don't hear from the brain washed sales force of that company.

So is Larrys announcements serious? Most likely these statements are made to reassure nervous SUN customers for as long as possible until Oracle figures out what to do with the SUN hardware. If he actually is serious IBM will of course not sit still and watch Oracle take a bite off its market. I think Larry will lose a lot in that fight.

IBM has DB2 but it runs mostly on z/OS and iSeries. UNIX is Oracle land, Windows is Oracle/MSSQL territory. DB2 is almost hidden, not very much marketed and if you want to try it out you have a hard time finding it on ibm.com. I had a go and couldn't locate it anywhere. Oracle DB however is two clicks away. No 60 day trial version. No crippled "free" versions. No complicated licensing scheme. No need to fill out any forms documenting everything about yourself and your company. Full versions at disposal if you agree not to use the products commercially. It is pure genious. People who know and has a feel for the product are much more likely to become customers.

Of course IBM could do the same and put DB2 up for a fight in the UNIX/Windows market. IBM doesn't rely on its database for its success but Oracle do. Pushing DB2 will hurt right in the core of Oracles business.

Oracle is a successfull company making a lot of money from database licensing. I think Larry will carefully analyze the risk of changing their ways.

#5 lparvirt on 2009-10-07 21:57 (Reply)

Disclaimer: I am an IBM employee who has been involved with the DB2 pureScale launch.

You asked if DB2 pureScale is a reaction to Oracle Exadata. I can tell you it's not. DB2 pureScale is actually the culmination of a multi-year collaboration between the IBM DB2 and IBM Power Systems teams. So the DB2 pureScale announcement was no short-term reaction to Exadata v2. If you don't believe me see:

http://it.toolbox.com/blogs/database-geek/db2-purescale-zos-envy-finally-over-34681

And, if you dig a little, you will see that Exadata v2 and DB2 pureScale are actually quite different. With Exadata v2, Oracle are looking to consolidate both OLTP and BI workloads in one environment. There is some good debate among well respected members of the database community as to whether this is a good idea or not, see:

http://www.dbms2.com/2009/09/29/integration-oltp-data-warehousing-exadata-2/

Instead DB2 pureScale is focused on making it easy to add capacity to OLTP systems (with true application transparency eliminating the need to make applications cluster-aware, and eliminating the issues that RAC has around "locality of reference"). DB2 pureScale also offers improved levels of availability and performance.

Thanks for the pointers about DB2 being difficult to find on ibm.com. You are correct. It could be better. I'll see what we can do.

As regards making DB2 accessible to the masses, we do have a freely-downloadable version of DB2 for Linux, Unix, Windows, and even Mac! And the free version has all core features, including pureXML. There are NO data limits with the free version. You can use it with as much data as you want. And there are NO time bombs. The only limits are that the software will use at most 2 CPU cores and 2GB of RAM. This compares favorably with Oracle who limit the amount of data and who have a lower limit for RAM.

But, you know what, I must say you are correct, it is difficult to find this free version of DB2. You can download it from:

http://www.ibm.com/software/data/db2/express/

#5.1 Conor O'Mahony (Homepage) on 2009-10-12 21:14 (Reply)

So this blog tells us that IBM and "PureScale" is a "Me too" response to Oracle!

Nonsense.

As far as "me too" goes, both Exadata1 and Exadata2 were Oracle's "me too" responses to Teradata, etc.

Get it??? EXA(data) vs. TERA(data)??? Even the marketing is "me too".

(BTW, Exa-BYTE was a tape-backup company from the 1990's).

The Exadata1 approach was necessary because Oracle was getting it's ass kicked in data warehousing by EVERYONE, no matter which hardware it runs on.

The Exadata2 approach is in reality nothing more than Exadata1 replacing HDD with Flash -- aka Oracle's "ExaFlashDance".

But this is also "me too", because HP and others have been doing this for almost a year now. In TPC-H, HP and Dell showed that they could get (almost) the world's best cost-per-query-per-hour (and kick Oracle's ass) with either Fusion-IO SSD on the PCIE bus or SATA SSD in JBOF.

The problem for Sunacle and ExaFlashDance2 is this:

Compared to ALL of the TPC-H-on-Flash-SSD results, Sun's MySql running on spinning-disk-only based (KickFire) hardware is:

(a) faster,
(b) 30-50% less expensive per QpH, and
(c) more energy efficient than TPC-H on Flash SSD

Unfortunately...Sunacle has no where to go here because Sun FMOD/SSDs cost 10x more per GByte and 5X more per IOP than the Fusion-IO SSDs that HP and Dell used.

Instead of betting it's future on being Ellison's "Flash SSD Oracle Accelerator puppy", Sun should be figuring out how MySQL on KickFire can be so much faster, cheaper and just as energy-efficient as Flash-based systems. (Hint--it's about KickFire's innovative new processor architecture, which is something Sun used to be good at.

Meanwhile, Oracle also gets it's ass kicked by Sun with Paraccel (in TPC-H), and has been for a long time now.

Funny thing...none of Sun's TPC-H results with Paraccel were ever challenged until just around the time Ellison made his play for Sun, when the world's record Sun/Paraccel "million-QpH" result at 30TB badly embarassed Oracle. It just so happens that prior to June of 2009, Oracle was the only player who ever submitted a 30TB result.

Not only was Sun/Paraccel 30TB result several times faster than Oracle, but it cost 95% less than Oracle -- 1/20th the cost-per-QpH -- and again it used no Flash.

The Sun/Paraccel 30TB result is now "withdrawn" just before we are supposed to see the first ExaFlash2 results. When these come out, we'll see that the cost-per-QpH of Oracle-on-ExaFlash is 10X higher than the Sun/Paraccel result on spinning disks.

And it's a pretty sure bet that we won't ever see any more Sun/Paraccel benchmarks....
http://paraccel.com/data_warehouse_blog/?p=104#more-104

And we won't see any more Sun MySQL results like this...

http://www.mysql.com/news-and-events/press-release/release_2008_13.html

#6 Steve A. (Homepage) on 2009-11-03 17:14 (Reply)

Add Comment

Name
Email
Homepage
In reply to
Comment	E-Mail addresses will not be displayed and will only be used for E-Mail notifications. To prevent automated Bots from commentspamming, please enter the string you see in the image below in the appropriate input box. Your comment will only be submitted if the strings match. Please ensure that your browser supports and accepts cookies, or your comment cannot be verified correctly. Enter the string from the spam-prevention image above: Standard emoticons like :-) and ;-) are converted to images. Enclosing asterisks marks text as bold (word), underscore are made via _word_.
	Remember Information? Subscribe to this entry