New Year’s High Def, Most Def
January 13, 2014 Timothy Prickett Morgan
Like many of you, I think New Year’s resolutions are suspect. While I think it is perfectly reasonable to stop and contemplate everything you have done and everything that you have yet to do after a certain period–the time it takes the Earth to make one circuit around the Sun is as good as any–I know enough about human nature to know that people do the best they can under the circumstances and that this often falls far short of the ideal. This is a good thing, I think, because the perfect can become the enemy of the good enough and stymie all change.
That said, I think that the IBM i community has much to look forward to in 2014 and that much of the change I suspect is coming to the Power Systems product line and its IBM i environment will be good for customers. But there is always something that can be done better, so to kick off the year for The Four Hundred, I will be mixing up a hearty stew of resolutions, prognostications, and admonitions for 2014.
The most obvious thing that will affect the Power Systems line this year, and therefore the IBM i platform, is the ongoing competitive pressure from X86 systems, particularly Xeon E5 and Xeon E7 platforms from Intel. Big Blue may still be the largest supplier of systems in the world, but Intel is the dominant maker of system chips, and it is almost as dominant in storage array processing and is working on being the biggest player in networking chips, too. IBM‘s 12-core Power8 processors will get integrated PCI-Express 3.0 peripheral controllers, bringing it on par with the latest Xeon chips from Intel, but the Xeon E5 chips have a dozen cores as well in the top bin parts and the Xeon E7s will have up to 15 cores. For commercial workloads, the Power chip has a clock speed, cache memory, and memory bandwidth advantage, allowing it to do more work, core for core and chip for chip. But Intel looks to be closing this gap, and that should be something that IBM is prepared to defend against.
The two-socket Xeon E5s are already out, having been launched in September last year, and the four-socket Xeon E5 chips will be out ahead of the Power8 chips, too. So IBM can actively tailor its Power8 system packaging and pricing to compete against machines using these Xeon E5 chips rather than having to reactively do so after the fact if the Power8 products came to market ahead of the Xeon E5 machines. Ditto for the “Ivy Bridge-EX” Xeon E7 v2 chips that are expected to launch sometime in the next few months and that are already shipping to server makers for high-end, fat memory X86 machines with four or eight sockets. (Or more, as in the case of Hewlett-Packard, which is tweaking its homegrown chipsets to make machines with 16 sockets using the Xeon E7 v2 processors. IBM could surprise us and do System x machines with eight and 16 sockets, too. Stranger things have happened.)
The initial Power8 machines come out around the middle of this year, most likely in midrange-class and enterprise-class machines if history is any guide, followed in short order by high-end boxes and then a refresh at the bottom of the line later in the year. The competition, including IBM’s own System x and PureFlex products, will be known and Big Blue will know what it needs to do to price Power8 machines to compete and win.
The question I have is this: What will IBM do in the meantime? Throughout the Power7 and Power7+ generations, IBM has shown little inclination to go aggressively after X86 workloads with either IBM i or AIX systems and instead has relied upon its lower-cost PowerLinux machines, which can only run Red Hat Enterprise Linux or SUSE Linux Enterprise Server and which cannot run IBM i or AIX, to win deals. IBM has put together bundles on PureFlex iron to make IBM i more attractive there, and there are some stale old trade-in deals that have been running so long I can’t remember when they were not. But that is not the same thing as making aggressive moves to try to get workloads on IBM i and AIX. I would think this would be a priority at IBM, and that the top brass would see the value of maintaining and extending the IBM i and AIX base. Making deals, very visible deals not one-offs that a sales rep does, to get both the vintage AIX and vintage OS/400 and i5/OS installed bases to move ahead should be priority one, and I expect for IBM to start aggressively engaging with these customers before they start engaging with its competitors, who mostly sell X86 iron these days. If anything, I would think that IBM would give the sweetest deals to customers who stay on Power Systems and then give competitive deals to customers who move to System x machinery using Xeon E5 or Xeon E7 processors. It is far better to keep customers–and keep them happy–than to lose them, knowing that it is ten times harder to get a new customer than to keep one and that once you lose a customer it is very hard to get them back at all.
It has been a mystery, to be honest, that there are not more competitive deals in the IBM sales book. I mean more in the sense of a greater number of deals as well as deals that really try to move customers ahead with deep discounts or other benefits. Maybe these deals exist and we don’t see them in weekly announcement letters, maybe they are done informally by business partners, or maybe IBM is just happy to accept what it is getting from the channel and to count on emerging markets and the top 3,000 or so of its accounts to make up the bulk of its Power Systems (and System z and probably System x) revenues.
It could be, of course, that IBM’s plan for a lot of Power Systems customers running the IBM i operating system is to get them off of systems running in their own data centers and out onto a number of clouds that support IBM i workloads. As I have pointed out before, it would only take about 1,000 enterprise-class Power8 systems to host the entire IBM i installed base. (These are beefy machines, to be sure, so that is in no way a slam.) But I think getting IBM i shops to go cloudy is going to take a little more effort and ingenuity.
A cloud is not just a hosted slice of a server. It is a way of automating the onboarding of applications from data centers to the cloud, hopefully using standard APIs and virtual machine instance types that can run on multiple clouds. Clouds also have automated means of expanding and contracting processor, memory, storage, and network capacity as needed. And most importantly, clouds let customers buy capacity quickly and fluidly and also allow them to turn it off as easily. A 12-month hosting contract with a monthly fee is not cloud computing. It is hosting.
For IBM i clouds in particular, then, I think what is necessary are a standard set of automation tools, most likely built on the combination of the OpenStack cloud controller and the PowerVM and KVM hypervisors, that would allow for workloads to be moved from private systems to public clouds more easily–and to allow applications and databases to be moved across public clouds, too. Someone needs to set these standards, and it probably should be the many cloud/hosting providers in the IBM i community working in conjunction with Big Blue so we get this done right the first time.
True cloud computing should come with true cloud pricing, which is generally available on a per-hour basis and, in some cases is available on a per-minute basis. Moreover, infrastructure cloud services for compute, storage, and networking need to be followed up with platform-level services, such as database, analytics, and other services. IBM i is a database service by default, of course, and IBM can work with software partners and cloud providers to get such services alongside basic infrastructure services. You have to offer customers the complete package, just as Amazon Web Services is doing and Microsoft Windows Azure, Google Compute Engine and App Engine, and Rackspace Cloud are all doing to one degree or another. I fully expect to see such infrastructure and platform cloud pairings for IBM i to emerge this year, and hopefully there will be plenty of variety and competition.
If you have predictions and hopes of your own for 2014 that you want to share, hit that Contact tab above and speak your mind.