MIPS (or MSU) is usefull as measurement on a Z/OS-platform, and cannot be used to compare with other platforms.
Imagine You have a diesel Truck with 500 HP and a petrol Ferrari with 500 HP - which one would you choose ?
You have to look at the type of workload and what the platform is designed for. The z-platform is excellent for massive amounts of data and transactions. It also is an all-inclusive platform == adding new workloads you just scale with PR/SM and manage with WLM, security (RACF/ACF2), crypto, back-up, disaster/recovery and much more is just there. The zOS-platform is a private cloud. Many have tried to move workloads from z to distributed and failed big-time. The savings aren't there. Many get blindfolded by looking at the $-numbers. For Z it looks as a big number, whereas the distributed has many distributed cost-elements that needs to be added together to get the true picture.
We recently upgraded our z13's to z15's, and went from model 7xx to model 6xx (subcapacity). With our workload we have seen a decrease in MIPS usage of +15%. This was achieved primarily through the effect of L1 cache.
Also take a look at this webinar (z and devops) https://www.linkedin.com/video/event/urn:li:ugcPost:6892086009874456576/
And this article https://jenselbaek.dk/zos/ - google translated https://jenselbaek-dk.translate.goog/zos/?_x_tr_sl=da&_x_tr_tl=en&_x_tr_hl=da&_x_tr_pto=wapp
------------------------------
Morten Løvschall
------------------------------