TomBascom
Curmudgeon
PROMON summary screen averages over extremely long are pretty much useless.
Even so -- your hit ratio sucks. 19,000 record reads/sec and 500 disk reads per second is a hit percentage of around 97% -- that stinks. 500+ disk reads/sec over all that time is pretty high for a 47 user system. But you have 9.5GB of RAM being used to buffer a 27GB db so something is churning your data pretty hard. I suspect a lot of table scans are going on.
Why are there 75 servers and only 47 users?
After image extents are defined but PROMON seems to think that after-imaging is not running. That is very, very bad. You are putting all of your data integrity eggs in someone else's basket. In my experience non-Progress persons cannot be trusted to do it right.
"Slow" is bad but corrupt or missing is much, much worse.
http://dbappraise.com/ppt/ai.pptx
Congratulations! You have type 2 storage areas. However, judging by the names, they seem to be setup "functionally" which is less than ideal: http://dbappraise.com/ppt/sos.pptx
If you require failover time of a few minutes then you should have a hot spare available. If you want latency on the order of seconds or minutes then you need OE Replication. That works fine with, or without virtualized servers. You certainly aren't required to be virtual to support such a scenario. Yes, you can do such things with VMs but you pay a price for that. Not only in dollars but also in complexity and possibly in performance. Especially if you are not knowledgeable about VMWare (or whatever virtualization technology you are using).
Even so -- your hit ratio sucks. 19,000 record reads/sec and 500 disk reads per second is a hit percentage of around 97% -- that stinks. 500+ disk reads/sec over all that time is pretty high for a 47 user system. But you have 9.5GB of RAM being used to buffer a 27GB db so something is churning your data pretty hard. I suspect a lot of table scans are going on.
Why are there 75 servers and only 47 users?
After image extents are defined but PROMON seems to think that after-imaging is not running. That is very, very bad. You are putting all of your data integrity eggs in someone else's basket. In my experience non-Progress persons cannot be trusted to do it right.
"Slow" is bad but corrupt or missing is much, much worse.
http://dbappraise.com/ppt/ai.pptx
Congratulations! You have type 2 storage areas. However, judging by the names, they seem to be setup "functionally" which is less than ideal: http://dbappraise.com/ppt/sos.pptx
If you require failover time of a few minutes then you should have a hot spare available. If you want latency on the order of seconds or minutes then you need OE Replication. That works fine with, or without virtualized servers. You certainly aren't required to be virtual to support such a scenario. Yes, you can do such things with VMs but you pay a price for that. Not only in dollars but also in complexity and possibly in performance. Especially if you are not knowledgeable about VMWare (or whatever virtualization technology you are using).