Mar 16 ’10

Linux on System z: Looks Like We Made It

by Editor in z/Journal

To a great degree, IBM’s power to endorse a technology governs what industry and CIO/CTOs say and do in the five- to 10-year cycle following a major strategic announcement. In this case, it’s gratifying to see IBM endorse a massive-scale virtualization product. The IBM Enterprise Linux Server (ELS) offering might just be one of those game-changing announcements, provided IBM can deliver the necessary services and application integration to make the package palatable to VMWare-crazed infrastructure engineers.

The Enterprise Linux Server for System z offering combines z10 hardware, z/VM, assorted IBM middleware, and a Linux distribution (sadly, only the “privileged two” distributions—Red Hat Enterprise Linux (RHEL) and Novell SUSE Linux Enterprise Server [SLES])—as a combined server consolidation and application farm. Marketing the original Test Plan Charlie argument of resource allocation, consolidated management, hardware reliability and application ubiquity for both large-  and medium-scale deployments, it’s an interesting attempt to attract both the traditional IBM markets and the new server hosting opportunities that are building from the groundswell of Linux support for general user applications. The pre-integration of the virtualization layer and the Linux infrastructure with z/VM Center is a long overdue step toward addressing the capability of a “zero knowledge” z/VM deployment without crippling the flexibility and functionality of z/VM. All well and good … so far.

After looking at the offering and talking to some people who’ve been using the prototypes, there are some questions regarding investment areas that IBM will need to answer to sustain a genuine success. The two biggest issues deal with sustained political will to counter the historical power of the z/OS organization: protection needed for the entrenched revenue stream that z/OS software represents, and the necessary investment to train and deploy sufficient services staff with z/VM and Linux skills across a wide area.

It’s safe to assume the z/OS organization is very concerned about its revenue stream. Linux for System z (and OpenSolaris for System z) challenges many assumptions about the right platform and right tooling to manage enterprise applications. By aggressively marketing Linux on System z, IBM is probably writing the swan song for z/OS’ dominance of the System z revenue stream and inserting the hardware into a much more competitive area against both Dell and HP. OpenSolaris even brings many of the enterprise improvements that Sun made to put UNIX in the same time zone for enterprise applications back to the System z platform, ahead of what Linux is delivering in some areas. In the past, the enormous revenue stream generated by z/OS has allowed that organization to drive much of the message and incentives for the IBM and partner sales force.  This offering opens the door to a system sale without any z/OS—heresy in many places in upstate New York. To be successful, IBM will have to confront this attitude head-on—a scary culture change that will need fundamental executive support from the top echelons of IBM to sustain. So far, the message has been clear, but the next three to four months will determine how much political will is needed to sustain the idea of a System z sale where more of the controlling software infrastructure doesn’t directly rely on z/OS. It will be interesting to see how this affects IBM spending on z/VM—buying real estate in Endicott might be a good idea if you’re a follower of IBM strategy.

The second concern—the difficulty of finding good z/VM and Linux talent for IBM’s services arms—will be a much longer-term battle. There are many very bright, very visible IBMers working hard to change this, but most are in the U.S., and most are so overwhelmed by the demand for skills that sustaining the deployment of the ELS offering will be a challenge. Partners will fill the gaps, but those partners also will need help in delivering—it’s still a trickle coming from the IBM educational co-op programs, and we’ll need more people than are available.

Those concerns are perennial in this space, so we’ll just have to see what happens. I’m just happy we’ve arrived at the level of prominence where some radical thinkers in IBM are willing to risk the crown jewels.

One interesting side item is the IUCV-based terminal server code introduced in the developer stream. This nifty new feature allows a virtual machine to attach to the console of a virtual machine as a pure ASCII device. If combined with the ASCII console function of the HMC, this provides a console KVM-like function, just like a serial concentrator and a traditional UNIX machine. I have some small complaints about the command syntax (why not simulate the standard Cisco-like syntax for terminal servers so all the management tools “just work”?), but it’s a great adaptation of the LAT prototype SNA discussed on the IBMVM mailing list. Support for the console function should be available in the major distributions in the next major release—have a look at it.