('binary' encoding is not supported, stored as-is)
As we can see from collect's output, that our system is totally running for the OS.
# CPU SUMMARY
# USER SYS IDLE WAIT INTR SYSC CS RUNQ AVG5 AVG30 AVG60 FORK VFORK
33 67 0 0 519 68187 7132 13 21.95 17.31 16.11 132.45 0.20
# SINGLE CPU STATISTICS
# CPU USER SYS IDLE WAIT
0 30 70 0 0
1 37 63 0 0
2 38 62 0 0
3 36 64 0 0
4 30 70 0 0
5 32 68 0 0
6 35 65 0 0
7 26 74 0 0
I have searched in the technical tru64 docs ( System Configurationa and Tunning) and Hp related this situation with 4 major possibilities.
- High number of system calls
- High number of interruptions
- High number of IPC calls
- High number of small IOS
Due to the nature of our application, it is more probable, that we have problems with the last one.
But how can I check this ( If I shall use the kernel debugger, how to use it) and say certainly that the problem is really this. And if there is some tunning that I can do in the OS to avoid this situation.
Hp technical manual also talks about a storage feature called "Write-Back Cache" that should be activated. That decreases the latency of small writes.
Someone has already got this situation before and can help me?
Email : pcoelho3_at_ibest.com.br
---
Inscreva-se na Maratona iBest para concorrer a prêmios!
São 2 casas, 11 Ford EcoSport e 60 computadores.
Participe já: http://maratona.ibest.com.br
Received on Wed Jul 07 2004 - 17:49:22 NZST