When we first looked at this, one thing that wasn't recommended when using a CP
versus an IFL was using CP Linux to run a DBMS. David, you helped us with the
analysis at the time.
The I/O intensive nature of the DBMS was a drawback, as was the licensing (you
had to license at the full MIPS of the entire machine versus the MIPS of the
IFL). Our non-DBMS deployment of Linux on CP has done its job beautifully,
though.
-----Original Message-----
From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu]on
Behalf Of David Boyes
Sent: Wednesday, October 07, 2009 9:28 AM
To: IBMVM@LISTSERV.UARK.EDU
Subject: Re: z/VM Linux on Cp
> Please explain, for those of us not yet involved in Linux, why
> it's not cost effective. For example, if we already have z/VM running,
> there is no additional cost involved.
If you already pay for VM on your standard engines, the issue is less serious.
But:
1) Standard engines are 4x (or more) the cost of an IFL for equivalent
horsepower.
2) Standard engines increase the computed size of the box for software
licensing purposes. This can be a killer for software in other LPARs,
particularly z/OS-based products.
3) Standard engines can be crippled; IFLs always run at full speed. Price per
MIP is much better on IFL, especially when you can get 4x horsepower for the
same spend as 1 standard CPU.
4) If you license VM on standard engines, you have to license it for all the
standard engines on the entire box, not just the IFLs in the LPAR in question.
5) Use of Linux on IFLs often reduces the need for other specialty engines that
are useless to general-purpose workload (eg, ZIIP, ZAAP, etc which are useless
to anything but z/OS). Better yield for same spend, and the IFLs benefit ALL
Linux workload, not just one z/OS instance.
It'll technically work fine; it just tends to not work out from the software
licensing and pricing perspective.
