August 21, 2005

zLinux TCO / Business case for zLinux

I've developing a business case for zLinux on the IBM Mainframe zSeries. From an initial glance it looks like the business case will be based on 3 things.

1) IFL - IBM's less expensive engine for Linux (still $125K) (you can move off apps from the more expensive mainframe engines)
2) Hipersockets - Virtual sesssions of linux can access zOS directly through memory mapped network access.
3) Linux Open standards - Besides running on z/VM linux is linux and anyone familiar with Linux or Unix can administer it. Unicode instead of EBCDIC, OpenSSL and standard open software and tools.
4) Less power usage
5) IBM Communication Controller for Linux on zSeries will allow for FEP removal

What to use it for?
IHS Apache
DB2 Connect
Tivoli
CCL

And potentially:
zWAS
DB2
Oracle
Lotus Domino

Limitations include:
Limited batch processing support, JCL is completely absent, and COBOL, is rare rare if used at all (OpenCobol supports a subset)
Security is more limited than RACF, and RACF integration is only via LDAP.
zVM will have overhead compared with running on bare hardware(5%?).
CPU is expensive for processor intensive compared with Intel/AMD
Spawning processes has more overhead on zSeries - there is a lot more checking and security then x86.

Backups?
FDRINSTANT
Tivoli Storage Manager

Monitoring?
Tivoli/mon

I am still looking at Suse vs Redhat for zSeries. Most people seem to be using Suse. Also according to Gartner group's Weiss Redhat support isn't that great as they have been growing too fast. I still personally lean towards Redhat as they are a "True" open source company. Novell sales still mostly come from close source apps (Novell/eDirectory etc).

Posted by Anthony at August 21, 2005 03:12 PM