One of the major issues IBM faces in the medium and long term with its new zSeries (Freeway) mainframe systems is the complete and utter lack of affordable development platforms for its Partners in Development. A proposed low-end 64-bit system - codenamed 'Flagstaff' - was cancelled in September 1999 and many have been waiting for a replacement to be announced. The [Numaq] is not it.

First of all, of course, it's only 31-bit. IBM's purpose in bring out a 31-bit system after already announcing the replacement (and backwards compatible) zArchitecture can only be guessed at - there are surely enough redundant 31-bit processors on the mark that could be recycled relatively cheaply. In any event, NUMA-Q does not provide the much-needed affordable low end 64-bit system without which the supply of applications from smaller partners will dry up. This is yet another cause for the loss of small applications from the platform - a process so damaging to application synergy over the last decade.

Secondly, it's not a System/390. It will only be sold by channel partners who are qualified for both NUMA-Q and System/390. Any System/390 partners who wish to sell it will first have to complete IBM's minimum certification for NUMA-Q - a non-trivial and not cheap business.

Thirdly - again - it's not a System/390. It doesn't even support ESCON peripherals or the current level of OS/390 (i.e., no Architecture Level Set 2 support). Does it support System 390's RAS features such as Alternate CP Recovery or Subsystem Storage Protect?

The way forward seems to be to revive the 'Flagstaff' project as a matter of urgency, even if this means it must be built by someone other than IBM.

"Emulating System/390 on Intel is like a dog's
walking on his hind legs. It is not done well,
but you are surprised that it is done at all."

Apologies to James Boswell - The Life of Samuel Johnson - 1791

Home Page