IBM To Launch New T-Rex Mainframe


News: IBM To Launch New T-Rex Mainframe

  1. IBM To Launch New T-Rex Mainframe (3 messages)

    In an effort to keep the mainframe alive IBM is set to release its next offering, the z990. Nicknamed "T-Rex," the system initially is expected to include 32 processors. The T-Rex will run Linux and J2EE 1.3 via Websphere. Mainframes are popular for server farm consolidation, allowing multiple separate server/OS instances to run on a single mainframe.

    Related articles:
    IBM To Launch New T-Rex Mainframe.
    IBM readies T-Rex mainframe for launch tomorrow .

    An interesting and prophetic article about the the application of mainframes and linux for server consolidation was published here on TSS in 2001 by Billy Newport (before he joined IBM):
    Corporate Data Centers, a soon to be extinct species?.
  2. I think that the mainframe is a very interesting capability for Java/J2EE.

    However, I wonder about the cost of these things - they are very expensive - and whether they will actually sell these things to "new" clients (ie those that are not usually accustomed to buying mainframe hardware).

    Their price limits them to only the very large companies. And even then, many of the largest companies arent structured in a way to budget for this magnitude of expense - traditionally, different divisions of a company are very reluctant to share costs.

    The idea of outsourcing is interesting... - but again, for many cultural, logistical or regulatory reasons, outsourcing wont be an option for many companies.

    Then, for those companies that are able to overcome the above issues, most would have to justify the cost by decomissioning numerous smaller, existing servers - which is a large and risky undertaking.

    So, while I really like the idea and while I realise that IBM are not planning for these to be a mass-consumption server, I wonder how many of these they will really end up selling?

    Especially with cheap x86 alternatives that provide much of the hardware virtualisation benefits of the mainframe - but for a much lower entry cost.

    The uptime and reliability of mainframes has traditionally been a big selling point for mainframes - yet how important is this uptime if you are regularly re-deploying your application to add functionality?

    Does anyone know of "non-mainframe" clients that are planning to consolidate servers onto them (ie buy new ones)?

  3. Running Java/J2EE on a mainframe is just too expensive period. Not only the expenditure on hardware, disk, tape and network components just to get it running but all the people needed to support that world. And the older the mainframe, the more people you have. T-Rex will help existing customers looking for mainframe server consolidation, data center footprint reduction and few people to support. IBM likes mainframes because of software renewals, lots of revenue from CICS. Most people will stay on mainframes because it's cheaper to keep it running vs. rewrite and deploy on cheaper hardware. Reliability is nice but mainframes go down for planned maintenance all the time just like Solaris or NT.

    Cheap x86 hardware will continue to rise to powers beyond mainframes (eg. grid, 64-bit chips, blade) but mainframes will always exist with their cobol apps, CICS and batch processing capabilities. I'm sure IBM finds a few new customers buying mainframes but it's a pretty steady business for them refreshing existing hardware environments.
  4. If anyone has a mainframe "going down all the time for periodic maintenance like NT", I suspect they have an organizational problem, not a mainframe problem.

    It's true a mainframe takes many more people to support it than one server..but it can be the equivalent of hundreds (or more) separate servers. (A typical mainframe indeed has several hundred things going at once - many programs acting as applications, served by DB servers, on-line servers, file servers, fall-back servers, etc.) In some situations, it appears (no personal experience here) that if a company has hundreds of Linux servers, it might be easier to support them virtually in-memory in an appropriate mainframe setup, as opposed to barns full of boxes & connections etc. etc. Again it all depends on individual circumstances - definitely easer for companies that do have applicable experience.