The Power Systems-IBM i Road Ahead
June 13, 2011 Timothy Prickett Morgan
The Four Hundred has spent the past several weeks going over what the top brass and head techies at Big Blue say about how the IBM i platform running atop Power Systems iron is doing both economically and technically. This week, we close out this series of articles with a discussion with Steve Will, the IBM i chief architect, concerning the near-term future of the platform, by which IBM means the time between now and the end of 2013.
Will went over a lot of this material at his IBM i: Today, Tomorrow & Beyond presentation at the COMMON tradeshow back in May, and like most of you out there in AS/400 Land, I was not able to attend that session. So Will was gracious enough to give me my own private briefing, and was very happy for us to disseminate the information to you.
Generally speaking, here are what Big Blue’s investment priorities for the IBM i platform are for the next few years:
“The top part of that chart is something you see all the time,” explains Will. “This is just table stakes, but it is important table stakes.” IBM owns RPG, so it pretty much sets the standard there, and as for COBOL, believe it or not there is an update of the COBOL standard that is in the works and due sometime this decade with more XML and object-oriented support. The core ISO/ANSI standard definition for SQL has 93 items, and all of them are implemented in DB2 for i 7.1, so it is hard to imagine what needs to be done here. The other DB2s from IBM have some work to do on the SQL standard front, however. IBM’s plan is to update DB2 for i once every six months or so, tweaking it to improve performance and add features.
In terms of making IBM i systems more resilient, Will says that IBM is trying to figure out how to simplify the installation and use of its own PowerHA clustering technology. The goal, says Will, is to get high availability in the hands of more customers, and that means making it easier for sure and probably making it less expensive in some ways as well. I think it requires both, and it probably means integrating with cloud-based HA and DR services that can run IBM i workloads and store and secure their data on an emergency basis.
IBM also needs to flesh out support for various storage area network switches and disk arrays as part of its effort to make HA and DR more pervasive, and to its credit, IBM has done a pretty good job of hooking in its various DS series of external disk arrays to the Power Systems-IBM i combo through the Virtual I/O Server in the past several years.
Simplified management goes hand-and-hand with every other theme of the AS/400 and its progeny. I have already told you at length about the new Systems Director Management Console that has replaced the HMC with the latest round of enhancements and also detailed further the transition from the HMC and IVM consoles to the SDMC with the future Power7+ and Power8 generation of machines.
Which brings me to an important point. Will did not talk about the Power7+ and Power8 processors in particular, but of course they are also in development. Before I even talked to Will, I told you all there is to know about these future processors and their platforms. Power7+ servers are due in the next year or so, perhaps in late 2011 or early 2012. Not much is known about the Power7+ chips and even less about the Power8 chips, but IBM is moving to an improved virtualized I/O scheme with the Power7+ chips that supports the SR-IOV standard that X64 chip makers have cooked up.
A lot of the work that IBM is now between now and 2013 has to do with making the IBM i platform more cloudy, as you might expect. Here’s what customers need, as far as IBM can tell:
What is ironic to me is that many of the features in the original AS/400–an integrated runtime, file system/database that exposed databases as services–are part of so-called platform clouds today. As if these ideas were new. IBM had a lot of the right ideas back in the 1970s, as it turns out. It was just that the hardware technology to drive the sophisticated software did not exist, and even if it did, it would have been insanely, inhumanly expensive. An integrated database with a simple, high-level programming language that keeps programmers focusing on business logic, not infrastructure, is what a System/38 was all about. It is also what Google is yammering about with App Engine and Microsoft is going on about with Windows Azure.
Anyway, those items above are the capabilities that IBM knows it needs to deliver in the IBM i platform to make it more cloudy. Now here’s the general roadmap for getting us there:
The suspend/release feature for PowerVM logical partitions shown in the future portion of the development roadmap above was actually delivered with the Technology Refresh of cumulative patches for IBM i 7.1 in May. The image provisioning and cloning tools, partition mobility, virtualization of more I/O features, and more automation are coming in future refreshes and whole versions.
“Even if customers don’t want to build a cloud, there are technologies that are part of a cloud-capable system that customers will want to take advantage of,” says Will. “For instance, image provisioning and cloning is the top cloud request we are getting from customers. They don’t want to go through an install every time they want to deploy a partition.”
I have been hammering on the fact that PowerVM needs to support live migration of running partitions, which is available for AIX on Power as well as for VMware ESX, Citrix Systems XenServer, Microsoft Hyper-V, and Red Hat KVM hypervisors. My attitude is simple: I think that anything good or interesting another platform can do, the IBM i platform should do. But Will says that when Big Blue talked to IBM i customers above partition mobility, which it calls live migration and which requires virtualization of system I/O, the largest customers said it would take them years to move to virtual I/O and storage. “This is why mobility is later for us,” Will explained.
I also think that it would be interesting for IBM to provide workload partitions, as it does for AIX. With workload partitions, there is one kernel and file system for a given set of logical machines, but there are sandboxes atop this single kernel and file system that allows each sandbox to look like a logically distinct machine, with their own security, authentication, and system and application software stacks. Solaris, HP-UX, and Linux all call these has containers; Solaris containers are mature, HP-UX containers less so, and Linux containers are still under development. It doesn’t look like IBM i will be getting them. Unless you make some noise about how IBM i subsystems are not sufficient.
Big Blue thinks it is making all the right calls on the IBM i development roadmap.
“So many of our customers have only one system today, so these features are not all that useful to them,” says Will. “But as they add a second system in an HA scenario, that will change.”
I would say it is a chicken-egg problem. If these capabilities had already been there, IBM might have sold a lot more second systems or fostered the creation of HA and DR clouds, which are now getting a toehold in the market despite the lack of workload partitions and live migration of logical partitions.