Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

There's a lot that you're misunderstanding here. A z mainframe can support a ton of VMs as long as the VMs are completely idle. A VMware cluster could do the same thing. The performance of z is so "good" that it will never be disclosed publicly.

But that has nothing to do with the cloud, because the cloud is all about automation. Mainframes are about manual provisioning and management. IBM does have experience with automated provisioning of HPC clusters, but that's also fairly different from the cloud.



> The performance of z is so "good" that it will never be disclosed publicly.

Is that true? I know nothing about mainframes, so it's an honest question


Maybe he's referring to the way the mainframes are sold with performance-limiting logic. When you want more performance, you ring IBM and they remotely activate more cores or upclock them.

IBM often writes "performance test" periods into their contracts -- you get x weeks per year to trial the higher levels of performance. This is a sop that realises that mainframes are busiest at financial reporting times, especially at the end of financial year.


No, I mean the relative performance and price/performance between z and Power or z and x86 is not public and never will be, so any performance comparisons between mainframes and normal servers have to be treated as rumor or speculation. This is not relevant if you're actually buying a mainframe because IBM will probably let you try before you buy, but it makes armchair architecture (HN's favorite sport) difficult.


In fairness to IBM, that's how enterprise sales usually works. Try finding independent production data on Oracle DB, for example.


Oracle's contracts expressly forbid talking about it unless Oracle have signed off on it.

Which is a pity, because I look after a bunch of Oracle on zLinux and would like to share...




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: