Hi folks,
I have a MTS issue if any of you may have dealt with before. We have a database with MTS and with 1200+ users. We initially setup 10 dispatchers and 10 shared servers, but all dispatchers got about 30% busy and consumed a lot of cpu. Each dispatcher used about 9% to 10% of cpu. The server is a 10-cpu 10G-RAM Sun 6500. To reduce contention on dispatchers, we added 30 more dispatchers (at slow increments) and now dispatchers are about 2% busy each sharing the load of 25 to 30 connections. CPU usage of dispatchers have reduced to less then 0.5% each which is good.

The problem now is, the contention seems to have moved to the shared server processes. The first 5 or 6 shared server processes (s0000 thru s0005) are now consuming most of the cpu, about 8% to 9% of cpu each. I found shared server terminations/creations upto 25, so I increased the initial server set to 40. However, the load is not getting distributed among the 40 servers, and it's still the same 5 initial servers that remain busy. As I understand, oracle should create/terminate new servers depending on load, but what I find here is that the load is not getting distributed among the shared server processes.

Any comments?

Thanks.
Saroj

Oracle LazyDBA home page