Does anyone have any experience in how best to configure processors on virtual servers? We have a client that upgraded their server and is now experiencing issues with some batch oriented programs (reports, updates) taking noticeably longer to run than on their old server. Their IT guy has run tests on both old and new servers and is showing stats that in most cases the new server is returning better results. The programs that seem to be running slower are doing a lot of reading and some rewriting and some report creation. The IT guy is asking if it makes any difference to the AcuCobol-GT Runtime as to how he configures the processors, such as 2 CPU's with 4 cores each, versus 1 CPU with 8 cores, etc. We do not do much multi-threading within our Cobol programs, and not at all in a few they are seeing decreased performance in, but not sure what/if any of this processor configuration on a VM matters to the Runtime either, or if it does affect it at all. Just curious if anyone had ever done any experimentation in this area. Thanks!
------------------------------
Wayne Epperly
Systems Analyst
Data Resources Corporation
Summerville SC US
------------------------------
- I'll let the Rocket people speak to how the runtime works with mutiple CPU/cores, but I think the cobol THREAD feature doesn't use them (it's an internal switch in the runtime). And I don't think the runtime process takes advantage of multiple cpu/core either.
- Did the old server have a faster drive? (One of our server upgrades a while back moved from 10,000 RPM to 7200 due to availability/cost. Some things sped up due to faster CPU and more RAM, but some slowed down significantly due to slower drive.)
------------------------------
Robert Redekop
Senior Programmer/Analyst
Samco Software Inc
Surrey CA
------------------------------
- I'll let the Rocket people speak to how the runtime works with mutiple CPU/cores, but I think the cobol THREAD feature doesn't use them (it's an internal switch in the runtime). And I don't think the runtime process takes advantage of multiple cpu/core either.
- Did the old server have a faster drive? (One of our server upgrades a while back moved from 10,000 RPM to 7200 due to availability/cost. Some things sped up due to faster CPU and more RAM, but some slowed down significantly due to slower drive.)
------------------------------
Robert Redekop
Senior Programmer/Analyst
Samco Software Inc
Surrey CA
------------------------------
Thank you Robert for your input. And I agree in that I think you are right about the Runtime not really taking advantage of multiple CPU's or cores. I wondered the very same thing about the disk drives since the delay shows up in programs that are heavy on file I/O. We ran into the same problem years back when a customer replaced their server and ended up going from like 10,000 RPM SAS drives down to 7,200 RPM SATA drives. Problem there was their IT company never asked for advice on hardware suggestions as we always stress to get the fastest disk drives you can afford. So then they had to end up replacing the disk drives with faster ones after all. Thankfully SSD drives have eliminated that to some extent versus the holder spinning hard drives. Before SSD's became so affordable we would try to get them to purchase 15K RPM SAS drives. But I have now got additional feedback from the IT guy in charge of this one that only Windows was upgraded from 2016 to 2022 and that all hardware is the same. So that makese this one even more puzzling.
------------------------------
Wayne Epperly
Systems Analyst
Data Resources Corporation
Summerville SC US
------------------------------
Thank you Robert for your input. And I agree in that I think you are right about the Runtime not really taking advantage of multiple CPU's or cores. I wondered the very same thing about the disk drives since the delay shows up in programs that are heavy on file I/O. We ran into the same problem years back when a customer replaced their server and ended up going from like 10,000 RPM SAS drives down to 7,200 RPM SATA drives. Problem there was their IT company never asked for advice on hardware suggestions as we always stress to get the fastest disk drives you can afford. So then they had to end up replacing the disk drives with faster ones after all. Thankfully SSD drives have eliminated that to some extent versus the holder spinning hard drives. Before SSD's became so affordable we would try to get them to purchase 15K RPM SAS drives. But I have now got additional feedback from the IT guy in charge of this one that only Windows was upgraded from 2016 to 2022 and that all hardware is the same. So that makese this one even more puzzling.
------------------------------
Wayne Epperly
Systems Analyst
Data Resources Corporation
Summerville SC US
------------------------------
Hi Wayne and Robert,
Thanks to both of you.
Robert is right, the runtime is not multi-threaded, so a single process won't benefit from multiple cores in terms of execution speed. Our multi-threading feature is an internal mechanism and does not leverage multi-core CPU capabilities.
To better understand the performance differences between the two servers, it would be great to use the internal profiler and generate a trace on both environments. This will help us pinpoint where the discrepancies are occurring.
Best regards,
Dominique
------------------------------
Dominique Sacre
Principal Product Manager
Rocket Internal - All Brands
Paris FR
------------------------------