Blogs

RedMonk

Skip to content

Can IBM POWER 7 hit the C-spot?

I attended a briefing last week in London with Rod Adkins, SVP IBM Systems and Technology Group, and Robert LeBlanc, who runs IBM’s middleware business. The subject at hand was the new POWER7 chip.

If you’re not a hardcore IBM server customer you’ve possibly never heard of POWER. It runs IBM’s mainframes, Unix boxes and the i (otherwise known as the most successful midrange platform and channel play in IBM history- the warhorse i like to still call the AS/400)

If you’re a hardcore gamer though you likely use POWER every day: Big Blue provides the silicon for XBox 360, Playstation3 and Nintendo Gamecube.

The question for POWER though is whether IBM can drive volume success with the chip in the mainstream IT market, which means database workloads. To win IBM needs to beat Oracle, which currently dominates the market sweet spot for mid to high end online transaction processing (OLTP) and decision support systems (DSS), a position it massively buttressed by expanding its application portfolio with numerous acquisitions. IBM DB2 has failed to take share from Oracle in the midrange Linux and Unix markets. That is IBM’s problem in a nutshell.

IBM maintains its high end market share leadership with DB2, CICS and IMS, but in IT volume is usually the charm. If IBM is right that Big Data and next generation analytics will be a bigger market than ERP it could be in a chance of some volume success. Clearly Big Data is a market, and systems inflexion point. Running smart grids, improving healthcare, instrumenting and monitoring a Smarter Planet.

But to win with POWER IBM will need to POWER the cloud, not just servers. The cloud will underpin Big Data, alongside the wave of 8 way Nehalem boards from Intel. That is – there is a double transition at work.

A few words on Fashion, a Lesson from History

Anyone that thought Oracle wasn’t serious about the Cloud because Larry Ellison dissed it just doesn’t know Oracle. I have been watching Oracle for 15 years now, so I have a number of Larry Ellison fashion t-shirts. Indeed Ellison is the Yves Saint Laurent of the tech world- never missing a trend. He talks it like Chanel, as if one little black dress would get you through life, but he sure knows how to play to industry hemlines.

I remember Ellison panning the notion of Grid Computing when everyone else started hyping it up. Same language – “its just fashion”. Then Oracle went ahead and called its next database Oracle 11g.

While IBM was architecting systems for customers that scaled to thousands of processors, Oracle, quite simply owned grid with that one little g. You see – not many people needed real grid computing. But they wanted 2 and 4 way clusters running Oracle RAC. And the g meant they could feel like they were doing that grid thing.

A couple of seasons before Ellison had dismissed the Internet, and then rolled out the Oracle 9i database. “i” is for Internet. See a pattern emerging? Oracle was always going to make a play for cloud marketing dominance, whether or not this was a tech “revolution” or not.

IBM needs to be careful not to confuse technical ambition, and the needs of a tiny percentage of the market, with the cloud C-spot.

POWERing the Cloud

What do Oracle and IBM both agree on? There is nothing new in IT. Ellison is even saying that with its acquisition of Sun his firm ready to deliver a promise like IBM System 360 mainframe.

But if IBM really wants to play in the Cloud sweetspot or C-spot its going to need to win more Web workloads. IBM should therefore worry less about SuSE and more about Ubuntu. What runs the Cloud? Ubuntu. What is the fastest growing OS for MySQL? Ubuntu. That’s the growth play. And Ubuntu can beat Solaris in volume – something no other IBM OS can claim. Hadoop on Ubuntu on POWER – that’s a Big Data play.

Virtualisation is somewhat of a wild card here, but unless we start to see Linux, Windows, and Solaris running as guests on IBM POWER-based servers, IBM’s play will be limited. VMware is currently better positioned here.

Weird Runnings.

Finally, I just want to talk some interesting installed base dynamics, namely that more IBM middleware runs on Solaris than any other OS. That’s right – Oracle servers and customers run the most IBM middleware. Like BEA (now Oracle), IBM knows hows to make money from Java. But the account management issues get more and more interesting. Oracle already supports many PeopleSoft customers running on Mainframe DB2 so all this is nothing new, but it should be remembered the next time someone from Oracle starts talking about cleaning and simplifying their stack.

In IBM’s favour will be intensifying competition between Oracle, Dell, and HP. In Oracle’s favour- account control.

I maintain that IBM blundered in not acquiring Sun, but now the bed is made. If IBM is to really hurt Oracle it must focus on volume, not the needs of a few global companies. It needs to nail the c-spot.

[update: please see the comments below for clarification on POWER family vs POWER chip. Thanks to Derek and Gordon Haff for reminding me not to let a good line obscure the bits. Also – I need to follow up with some comments on the other leg of IBM’s POWER systems strategy – appliances)

disclosure: Canonical, Dell and IBM are clients.

Categories: Uncategorized.

Tags: , , , , , ,

Comment Feed

28 Responses

  1. ネットを使った非常識なお金の稼ぎ方  http://sfurl.biz/k0sqN 100210164039
    This comment was originally posted on Twitter

  2. James Governor’s Monkchips » Can IBM POWER 7 hit the C-spot? http://bit.ly/do5MWL
    This comment was originally posted on Twitter

  3. Can IBM POWER 7 hit the C-spot? http://bit.ly/do5MWL /via @monkchips (I remember the i Series first as Project Silverlake ….;’D )
    This comment was originally posted on Twitter

  4. good post from @monkchips: Can IBM POWER 7 hit the C-spot? http://bit.ly/aEL0Du
    This comment was originally posted on Twitter

  5. re. “It runs IBM’s mainframes” … nope, apples and oranges do not make an instruction set (370) with little and big endians and applications.

    re. “XBox 360, Playstation3 and Nintendo Gamecube” … PS3 is a master POWER (not 7) with vector processors (cell processor) … also, XBOX360 … Nintendo is Broadway POWER (not 7) … still apples and oranges

    POWER7 is not a commodity, thus no cloud possible.

    Migrating DB2, CICS, IMS to cloud is huge (COBOL) … thus, they will keep the dino and shrink over time. Again, mixing POWER7 with mainframes is apples and oranges. Here is my 2c http://bit.ly/3JPUFI

    Oracle got it (i, g and now maybe c). IBM just had meetings (I was there for both i and g).

    Yah, they (IBM) definitely can’t kick the SUSE habit.

    Yup, no Windows and Solaris is a problem. The middleware on SUN will shrink out to the cloud anyhow as alternative from open source.

    Frankly, IBM does not get the cloud anyhow.

    Watch out for the kool aid!

  6. Hey Derek – thanks for that.

    you are right that the Cell chip, which is found in the games consoles is not the Power 7. The chip for z is also a relative, rather than a singleton of the POWER 7 chip. I was trying to make a point about the POWER family, and should have been clearer.

    Given how much IBM middleware runs on Solaris I do see the OS as significant. Ans I wouldn’t underestimate Oracle’s intentions for Solaris either.

    “Frankly IBM does not get the cloud” is a pretty broad statement. The fact you can purchase IBM middleware on Amazon Web Services is a clear indication of IBM pragmatism in the area. It has a number of vectors to address the cloud.

    POWER as yet, doesn’t seem to be one of them. Would anyone other than IBM choose to run a public cloud on POWER? Would a startup target the environment? If they were an analytics focused one they just might.

    Coolaid. People have been writing off IBM systems architectures for a long time. POWER is very much alive and kicking.

    And regarding Cell, mentioned above, some of the secret sauce in POWER7 which i didn’t talk to much in the short post above is based on shared architecture with the rest of the family.

    Enjoyed your post by the way Derek. Regarding real horizontal cloud architectures you should check out what Twitter is using using BitTorrent for – software distribution on its servers. http://torrentfreak.com/twitter-uses-bittorrent-for-server-deployment-100210/

    James GovernorFebruary 18, 2010 @ 4:16 pmReply
  7. POWER7 is perfectly suitable for PaaS and SaaS clouds, but IBM haven’t taken the virtualisation management up to a level suitable for running large-scale IaaS on POWER7 – the LPAR+VIO management still needs too much CLI knowledge of things like how to attach virtual disks to a server, or even simple things like attaching a virtual CD-ROM to the server to boot an ISO. The API layer for the management of this is minimal at best, and without a lot of improvement noone’s going to invest the time in building their own front-end to it.

    If you’ve got an application which can treat each POWER7 box as a powerful but disposable chunk of CPU resource, like a parallel database, then you’ve got a great building block for a “Data Warehouse as a Service”, “Message Queue as as Service”, things like that.

    Indeed I think IBM are working hard on data warehouses and “business analytics as a service”, though they call it mad things like “IBM Information on Demand”, which I’m sure will run on AIX and thus POWER7 when it gets rolled out.

  8. As you’ve subsequently commented James, Power 7 has some shared heritage with the cores used in xbox, playstation and to a point, mainframes – think Itanic and x86…

    I’m of the view that its really not that interesting anymore, IBM continue to tilt at windmills with it’s large central servers like System p and mainframes, its a battle no longer worth fighting. While there is a case to be made, and we did for Power 7, that going to a highly threaded structure is mostly redundant as the software just isn’t there to use it, creating a centralized behemoth just to run virtualized is the only game in town to keep the price up.

    However, that really isn’t where the business is anymore, and will be in the future. Intel x86 has way passed the good enough test. The software has matured to a point where it doesn’t crash nearly enough to warrant the complexity that has been built in over the past 5-years.

    For a more interesting approach to IBM’s computing future, go look at the work that Charlie Johnson in Rochester is doing on IBM’s A2 Core http://www.theregister.co.uk/2010/02/09/ibm_wire_speed_processor/ just don’t jump to the conclusion that because this is also related to the Power processor that it will find its way into System p/p-Series anytime soon. As you know, software compatibility rules uber alles.

    On that subject and given you raise the issue of Solaris, did you ask IBM what they’ve done with Transitive since buying them?

  9. Thanks Ewan. Great context.

    Mark you might want to disclose your affiliations. you do of course know POWER as well as anyone. Dell is certainly selling more servers to web and cloud companies than IBM. or Sun/Oracle for that matter. Arguing against central servers though? In the age of the cloud?

    James GovernorFebruary 18, 2010 @ 4:55 pmReply
  10. Ewan, good comments. I’ve tried finding any useful detail on the integration of IBM Director and the Power HMC in the Power 7 announcement, but alas couldn’t find any.

    They had a great opportunity to work through the simplification of the config, deployment, management and monitoring, as well as to support the evolution of LPAR and AIX WPARs as a cloud-like infrastructure with Power 7, by making this seamless, consistent etc. I hope they didn’t blow it..

  11. My affiliations are clear James, just look at the blog linked by my comments. Transparency rocks!

    As I understand it, and I may be wrong, Dell’s cloud hardware custom solutions unit would be the 3rd largest server vendor in the US, by volume shipments, if it were a standalone business.

    And while you could argue that these are centralized servers, especially where shipped self contained in a container, they are not from a software perspective, nor are they from a virtualization perspective. Unlike say a large centralized Power or mainframe server. And you know, that while I will defend hardware architectures, I’ve always been about the software architecture.

    So I’m not arguing against centralized, as you well know, I’ve always believed that this IS the right model, even in the client/server days; I’m of the opinion that monolithic is no longer the right model. To be honest, while IBM might have “8 way Nehalem boards from Intel” in their sights, that is and will continue to be a less and less interesting academic debate.

    How for example do you see the NOSQL, MYSQL and MAPREDUCE style implementations working on 8 way of any architecture ??

  12. #Cisco #Cloud James Governor’s Monkchips » Can IBM POWER 7 hit the C-spot?: Adobe Amazon android APIs AppEng… http://twurl.nl/dpli4b #TCN
    This comment was originally posted on Twitter

  13. #Cisco #Cloud James Governor’s Monkchips » Can IBM POWER 7 hit the C-spot?: Adobe Amazon android APIs AppEng… http://twurl.nl/4wtjro #TCN
    This comment was originally posted on Twitter

  14. James Governor’s Monkchips » Can IBM POWER 7 hit the C-spot? http://bit.ly/bJoj2M
    This comment was originally posted on Twitter

  15. James Governor’s Monkchips » Can IBM POWER 7 hit the C-spot?: I have been watching Oracle for 15 years now, so I h… http://bit.ly/au2pHP
    This comment was originally posted on Twitter

  16. James Governor’s Monkchips » Can IBM POWER 7 hit the C-spot?: I have been watching Oracle for 15 years now, so I h… http://bit.ly/au2pHP
    This comment was originally posted on Twitter

  17. James Governor’s Monkchips » Can IBM POWER 7 hit the C-spot? http://bit.ly/bsBQvV
    This comment was originally posted on Twitter

  18. James, I am not sure that the definition of processor chip defines what a cloud is or how a cloud works.

    After all, the ultimate “business” goal of a cloud is a virtual infrastructure that is in the form of a service, an application, actual infrastructure to build upon, or storage as a new form of cloud offering.

    The reality is that very few are offering a real cloud solution of any kind other than SaaS which is still very relevant but not as sexy to talk tech about these days. So instead we talk about cloud.

    Which reminds me that when talking about cloud, it harkens back to my very early days and using multiplexed time as no one could really afford their own computer system?

    Now my disclosures;
    1) I am an IBM client
    2) I am running the first Power 7 system in Canada
    3) I am a client of Redmonk
    4) I provide customer testimonials for IBM on Power, Linux, & Virtualization
    5) I am a Novell Enterprise Linux (Suse) client and customer reference on Power

    So there are LOT of items touched on in this article, my take on power as a processor.

    The Power 7 processor brings new processing power and speed to the table for IBM Power customers. Keeping in mind that HP, Sun/Oracle, and Dell – CAN NOT run 3 different operating system simultaneously in their virtualized environments. Even VMware only handles Windows and Linux workloads. The number of unique operating systems speaks to number of solutions available for a business to choose from that leverages the hardware investment. In other words it’s about software choice without infrastructure costs. IBM wins hands down.

    Now for the Oracle part. Larry has always been a flash in the pan kind of guy. Saying or doing something to get him front page news. Now he wants to rule the world and adopt IBM’s 1960’s view on being a total solution provider for his customers. IBM learnt the hard way that customers want choice, not a good direction for Larry. The reality to all this posturing is that Oracle enjoys very high end mission critical respect for its database and solutions, which by the way generates a lot of mtce revenue that Larry is desperate to protect. He believes that by owning the whole process he can guarantee his clients the most reliable experience that he controls. History tell’s us that unless choice is available in hardware, operating systems, and support they are destined to fail. Just look to players like WANG who dominated the early imaging market as one of many stories that tell this same tale. I wont even get on my soap box to talk about Larry’s Linux as that is a whole series of issues behind it. In the end if all you do is Oracle, then a 100% Oracle solution may be best for you, but reality is no one is 100% any one product, let alone just one product.

    A few words from a business perspective on database. When you run a lean shop which is the reality for all of us today, having technologies that are self healing, self managing that don’t require a DBA are of value. That is what DB2 does for us. As I understand it that is not the world of many database providers including Oracle. It is rare to find DBA’s in a DB2 shop, especially in an SMB space, I can’t say that for my counterparts that run other main stream database technologies.

    As far as powering the cloud, if it’s about choice then IBM has a unique offering in the Power system, but again I don’t see cloud as the end goal for a processor. The management tools for ALL vendors are not what we want today, that speaks to your point about creating front end tools in order to offer a service. I am not sure that business is really ready to jump to a cloud, so this posturing is about being ready, not what is ready today. That includes the way you can manage a cloud or offer services from the cloud.

    James knows that I push the Power envelope so Unbuntu Vs Debian Vs Red Hat Vs Novell Suse is again about choice, they all run Linux on a Power platform. I always stress caution to an enterprise audience about Linux in that if we are to learn from the past and adoption of Unix, the world of Linux must learn to not become to proprietary, which in the end is my biggest beef with Larry’s Linux.

    In summary:
    > A chip technology does not define the cloud
    > Power 7 is just the next generation in a roadmap of chip technology, guess what’s next Power 7+ and then Power 8
    > Power systems are a virtualization engine which is about choice for business
    > Oracle may be at risk of alienating it’s clients by eliminating choice
    > Running lean means deploying technologies that self manage
    > The cloud is a great concept but not widely deployed other than SaaS
    > Business is not ready to adopt the cloud
    > Nobody is managing the cloud really well yet
    > We have to be wary about splintering Linux the way Unix was

    Thanks for listening . . .

  19. Interesting stuff Nigel, lots of good info.

    One thing though, the vSphere OS compatibility is pretty good, you can run Windows, Linux, Sco unixware, solaris, freebsd under it, though some are experimental

    http://www.vmware.com/support/vsphere4/doc/vsp_40_new_feat.html

    As much as I like power + aix, it was my first it job after university, and I still work with it daily, the odds of power’s mainstream survival are just getting longer and longer.

  20. “What runs the Cloud? Ubuntu. What is the fastest growing OS for MySQL? #Ubuntu” @monkchips in http://bit.ly/bBaeBH #ubuntucloud #cloud
    This comment was originally posted on Twitter

  21. .@monkchips “What runs the Cloud? Ubuntu. What is the fastest growing OS for MySQL? Ubuntu. That’s the growth play” http://bit.ly/9UkoyO
    This comment was originally posted on Twitter

  22. #Cisco #Cloud James Governor’s Monkchips » Can IBM POWER 7 hit the C-spot?: #Cisco #Cloud James Governor’s Mon… http://bit.ly/au2pHP #TCN
    This comment was originally posted on Twitter

  23. #Cisco #Cloud James Governor’s Monkchips » Can IBM POWER 7 hit the C-spot?: #Cisco #Cloud James Governor’s Mon… http://bit.ly/au2pHP #TCN
    This comment was originally posted on Twitter

  24. @monkchips : excellent article – http://bit.ly/aEL0Du – spot on in my view with Oracle, the acquisition of Sun was all about private cloud.
    This comment was originally posted on Twitter

  25. So is IBM going to buy Amazon.com and spin off the book-selling business?

    Like you recommended what, three years ago?

  26. Finally I get around to this post (!).

    Nice commentary with the usual long-term perspective I’ve come to expect, James.

    As you know I’m an IBMer but I don’t speak for the company’s strategy (I’m a long way from that!). I’m interested in your comments about Ubuntu’s growth here – are there stats available? You already know that the IBM Client for Smart Work is going to be supported on Ubuntu but with the growth of Ubuntu Linux in the server space it’s probably worth watching what happens around server-side stuff as well.

  27. Don’t let the facts get in the way of a good blog entry. Oracle 8i was the first i database and it came out in 1995 around the time Oracle released a their thin client or Network Computer as they were referred to in those days. Hardly seems dismissive to me.



Some HTML is OK

or, reply to this post via trackback.