Techno Musings

Just this past week IBM announced their long awaited refresh to the zSeries: zEnterprise or z196.  The tech-media and blog-o-sphere will and has focused on the large changes in the system architecture, processor, software, and linkage between BladeCenter and the zSeries: zBX and URM (Unified Resource Management).  To be clear this appears to be IBM’s Exadata/VBlock killer, which they call an Ensemble.  However IBM’s interesting twist comes from their poly-system combining Mainframe systems, POWER7 systems and xSeries systems into a single managed whole.

Image attributed to Michael Cote

IBM’s approach seems to be a different take on vertical integration, and at least to me IBM seems to imply the promise of applications automatically being distributed across or taking advantage of these resources.  However, after reading IBM’s draft Redbook on Unified Resource Management (see chapter 13) what appears to be supported today is clearly spelled out as: workload management across resources, which is seemingly tied to goal seeking for performance and energy consumption objectives.  I would like to see more though and I think there is quite a bit of R&D to get there.

For sure IBM, NVidia, Microsoft, Apple, Cray and many others are and have been experimenting on a variety of technologies that take advantage of many-core or hybrid compute architectures.  (I’ve blogged quite a bit on the topic of hybrid computing, and you can find a smattering of different posts here.) The question becomes how can the R&D be applied to real systems for real users to take advantage of?  Personally, I hope that it can be done over time with the goal of minimizing human interaction.  After all we all have a sense that technology is deployed to help people get their jobs done and not for an end in and of itself. For example here is a quote from one of our customers, NASA, as to why they selected the Hitachi Content Platform.

“The biggest benefit of the Hitachi archive is knowing that we give it data and it gives the data back when and how we need it. But the fact that it is so easy to manage is a really great value to the organization because scientists can stay focused on the health of our planet,” concludes Curt Tilmes (Computer Scientist for NASA).

As Curt states the selection of HCP was about letting scientists do their job of determining the health of our planet, and not about having them operate and maintain the system.  Therefore, the goal of any new R&D activity in the industry must be to free the business owner, the scientist, the author, etc. from the burdens of operating complex systems.  As such R&D around advanced schedulers, operating systems, compliers, IDEs, management suites, etc. that help developers and administrators maximize utilization of their IT resources on hybrid architectures must be key in the future to continue realizing this point.  But — ain’t there always a but — getting the developers and administrators comfortable with the usage and deployment of technologies and applications which assume hybrid architectures and automate things significantly is another major part of the puzzle.  Without their buy-in and engagement all of the R&D in the world which promises to solve IT world hunger and fly the Space Shuttle at the same time will become shelf-ware.  In short both parts of the “equation need to be balanced” to get the righteous deployment of cutting edge technology.  Here’s to balancing the equation and realizing hybrid platforms for real!