top | item 24222678

(no title)

Misteur-Z | 5 years ago

IBM missed the 21st century turn, and buying Red Hat will probably not save everything [1].

Mainframes are just too expensive, even with open-source software running from the Canonical or Red Hat ecosystem. You just can't keep billing CPU consumption and sell memory sticks like it's made of diamonds. Running heavy "modern" workloads will give you a crazy stupid high R4HA MSU bills ... the clash between the relatively cheap commodity hardware philosophy of the last 10 years and the IBM way of billing high-end things is too big.

I would say mainframes are still really good at handling an ultra large bus with sky-high bandwidth but that's pretty much it. The power platform is a lot cheaper and also good at that.

[1] Earlier this year: IBM’s Lost Decade https://news.ycombinator.com/item?id=22224782

discuss

lsllc|5 years ago

IBM should stop trying to copy AWS/Azure/GCP with their cloud offerings and instead offered a "mainframe" based cloud, they might get more traction. Customers could submit their COBOL/CICS jobs and just pay for the cpu milliseconds or I/O blocks used etc. getting the cloud scale and elasticity for a mainframe without having to actually "own" a mainframe. I guess that's where AWS is going with some of their ML & analytic offerings (batch oriented, but not mainframe based).

lboc|5 years ago

The 'service bureau' does what you're talking about, and has done for a very long time.

https://en.wikipedia.org/wiki/Service_bureau#Histories

I have no idea about the current level of activity in this sector (mainframe proceesing as a service), but would not be at all surprised to find it greatly diminished from it's heyday when dinosaurs roamed the raised floor.

I don't think it takes a huge leap to draw a connection between these businesses and the cloud providers of today.

bobbydreamer|5 years ago

IBM cloud for learners itself not cheap that's the problem. But the thing is they cannot do it like that each software takes a huge load on the lpar/Plex and run as a subsystem. So cost of running a small stack itself will be huge. Now they got zVM don't know how effective that is, but currently it helps Devops process.

Generally Billing is in peak MIPS.