Dec 14 ’11
Linux on System z: Dear Santa …
With the holidays quickly approaching, the neighborhood kids have been making wish lists for Christmas and Chanukah, trying to maneuver around their parents for what the really cool thing of the year seems to be. I don’t know what that might be, as I have cats, not kids, but in that spirit, if IBM were taking requests, here’s my letter to Santa at IBM:
Dear Santa at IBM,
I’ve been a reasonably good engineer this year. I would like to ask for a few things that (if you’re so inclined) would make my holidays bright. First, I would like to see Unionfs (www.unionfs.net) become a standard part of both major Linux distributions. If you haven’t been following it—we know, elf industrial actions, reindeer veterinary trips, all that jazz—Unionfs builds a stackable unification file system that can appear to merge the contents of several directories (branches), while keeping their physical content separate. Unionfs allows any mix of read-only and read-write branches, as well as insertion and deletion of branches anywhere in the fan-out, and handles elimination of duplicates, partial-error conditions, and more. For System z, this allows overlaying parts of read-only file systems (delivered either by shared minidisk or DCSS) with writable data, which solves many of the problems with sharing large directories such as /usr between virtual machines but still allowing individual RPM databases and other customizations built on top of the shared read-only copy. It has a few quirks of its own, but it’s an interesting way to help build and deliver on the promise of shared resource machines.
My second wish,Santa, is for Layer 2 networking support on real hipersockets. Given the amount of pain and complexity involved in getting a hipersocket connection to work, the ability to use Virtual LANs (VLANs) natively on hipersockets would be a great improvement in the whole setup. I could define connectivity to other LPARs using normal networking architectures, and adding hardware acceleration (such as moving network switching to a SAP-like other system I/O) would be a huge performance and usability improvement—it might even make me change my mind on the pain/performance scale for hipersockets overall.
Third, if you’re still listening, I’d like to see 40G and 100G Ethernet OSA cards. Competitively, we’re going to need those speeds very soon as we continue to scale up virtualized system environments and those z/OS elves continue to improve database and batch processing environments to store more and more data. Coincidentally, converging data transport and storage area networking also need those types of speeds at scale, so you would really be doing us a favor if you expedited the arrival of the interfaces.
Last, how about CMS Pipelines support for SPXTAPE? A lot of us don’t have tapes on our VM or Linux LPARs, but it’s been really handy to have that support for DDR so we can work around that problem. Without SPXTAPE, we can’t dump our spool volumes on VM, so migration is really hard.
Santa, I know you’re trying. We don’t ask for much—just peace and goodwill. And new features. And hardware.
The VM and Linux Community
P.S. The cookies were supplied by the Dark Side. I’d be careful with them if I were you.
Best wishes, everyone. May all your futures be bright.