On 17/07/2021 8:46 pm, Brian Westerman wrote:
once they figured out that the two "sample" sites I was dealing with were z13s (80mip) 
and a z15t02 (88mip), they decided that they needed to republish the "minimum" 
requirements.  If you look at the guide now, they use a 400+mip box to do the setup and testing.
Software pricing has really strangled the small end of the mainframe market.

Mainframe workloads are usually distant enough from other systems that we don't properly see the scale, but the reality is that these systems are tiny by current standards.

I didn't realize the problem until I looked properly at the LSPR and CPU charts while investigating a customer performance issue.

I develop in Java, and it has been a useful coincidence that work runs at a similar speed on my PC and on the z/OS test system. That makes my PC with 6 cores roughly equivalent to a 8561-606 at 6000 MIPS. My PC seems about 10x faster than a Raspberry Pi, which makes a Raspberry Pi (4 core) equivalent to about 400 MIPS. These sub-100 MIPS systems seem to be roughly equivalent in CPU power to 1/4 of a Raspberry Pi.

This causes a number of problems:
- Customers spend a large amount of time managing performance
- It is impractical to move workload to the mainframe, and rewards moving workload off - It is a headache for software vendors that need to cater for very small systems - as IBM is discovering with z/OSMF

I would like to see IBM set a minimum system size more in line with current expectations - something equivalent to e.g. an Intel Core i3: minimum of 4 CPUs and 3-4000 MIPS.

Software vendors would have to adjust their pricing to avoid ridiculous upgrade charges, but overall it would make the mainframe healthier. Usage based pricing doesn't help either. It's no use having the capacity if you can't afford to use it.

--
Andrew Rowley
Black Hill Software

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to