V=R - then why run under VM at all? As an old MVS production hand, I have
to at least smirk at that ;) Why not VM for those things appropriate to VM
and an LPAR or LPARs where appropriate, talking back and forth using
hipersockets?

One of the configurations I see tried more and more often is a VM
environment of a lot of small linux EC's to act as web servers and routers
connected to one very large data base server (VLDB) (Linux LPAR,  MVS LPAR
or pSeries). Its these VLDB servers I'm talking about.

 These VLDB servers under Linux could use 2-10 GB of real memory and have
hundred's of GB's or Terabytes of data at 15 ms access times or less. In
these cases, why incur any VM? The apps themselves are large and the big
buffer caches for Linux speeds up the I/O.  This kind of app, is well
suited for zSeries parallel escon/ficon channel architecture is and, if
anything, I/O bound. Forcing the buffer space small only slows them down
and using large memory under VM only adds unnecessary paging. Things like
very large, frequently used table spaces and indices do much better in
memory.

Obviously, the issues of cost, performance, throughput, availability, and
reliability, need to be stewed into the equation, but don't discount BIG
penguins as an alternative.

Regards, Jim
Linux S/390-zSeries Support, SEEL, IBM Silicon Valley Labs
t/l 543-4021, 408-463-4021, [EMAIL PROTECTED]
*** Grace Happens ***

Reply via email to