Well we have an old, new mainframe again. I spent a lot of my time with mainframes but moved on when I saw the mainframe market going down, I also felt that I had done as much as I could so I moved to working with Unix, specifically Solaris and HP-UX.

Now there has been a large amount of FUD about this announcement. Some thing that I would like to put straight.

Consolidation - The Mainframe is good to consolidate other mainframes but not really other servers. So if you have 2 or 3 old mainframes that you cannot get rid of. Consolidate them into one of these new ones. Make sure that IBM can show that it will be cheaper to run one mainframe than 2 or 3. Then check the numbers very carefully and do your own research. About consolidating Unix (Linux) servers into a mainframe, now that is a big leap of faith. People should be very careful with this proposition. Should not IBM get more applications developed for z10 and zOS, Open Source z/OS to get better mainframe adoption. This IBM linux consolidation sounds like a desparate cry of help for the mainframe.

IBM mainframe consolidation problem number 1, if the z10 system needs to be serviced including an outage, you will have all your Linux apps out of action. Now IBM will then say buy 2 mainframes, oh dear. Best to stay where you are and consolidate in smaller more manageable and serviceable chunks.

Mainframe big/fast box - I do not believe this for one moment. The large Sun and IBM Unix servers are faster and probably a lot smaller than this old/new iron.

Reliability - The mainframe lost this years ago, Sun and other High end Unix mainframes/servers are just as reliable. The mainframe may be more reliable as you only run your old apps on them which you never change as no-one understands them anymore. But then which server would not be reliable, if you never interfere with it. There are many examples of Sun servers running for several years without a reboot and departments forgeting about them as they "just run forever".

Scaleability - Sun and other large mainframes/Unix servers scale to equivalent or larger systems. Nothing new here for the z10.

Performance - The MIPS algorithm or Meaningless Indication of Performance. Is really a red-herring, you can compare a old mainframe to an new, old mainframe using MIPS but it is not really useful for comparing mainframes with other server platforms. TPC-D type benchmarks are better than MIPS, but still is not a definitive comparison. Companies do not make money running TPC benchmarks they make money running applications, which can only partially be compared using benchmarks.

Innovation - Presently from the information that I have available, the new Z large server (aka mainframe) uses 2 dual core P6 CPU's on a module/circuit board to make it look like a 4 core system. Well you can fool some of the people most of the the time but not all of the people all of the time.

I see that I will need to spend some time looking at this mainframe stuff in detail.