Huge load on Database server

Question

Hello all,
Good day. We are using oracle standard edtion 9.2.0. We have sun machine with dual cpu and have 8G memory. We use decicated server for our web applications. We also use SharePlex for replication. The average number of logins (users) in the system is around 25-30. Suddenly we are facing tremendous load avarage on db server - highest is 20 that means 20*100(%).
In our daily load avarage statistics, we always find some peaks telling that most of the times, the load average remain very high. We have tuned our expensive queries that we found by sp report. For your kind information, the most expensive query do not take more than 2 sec when it runs individually. During a heavy load on db server, even a simple query (usually takes 200 ms) is found as a guilty party.
We have set a large value (1024M) for our pga_aggregate_target parameter for testing some queries that were using disk sort operations. Has it an impact?
Someone of us, run a 20 minute sampling of system statistics when the system was pegged at a load average of 20, and then flushed the shared pool which forced the existing sql to get re-parsed on the next execution. The load average on the system started dropping within seconds of doing this, and then went all the way down to a level under 1 in a matter of minutes.
What more we can do? Waiting for your helping hands :)
Here I'm providing some parameters for your kind considerations.
db_block_size 8192
db_cache_size 1073741824
db_file_multiblock_read_count 16
db_keep_cache_size 117440512
db_recycle_cache_size 117440512
fast_start_mttr_target 0
hash_join_enabled TRUE
java_pool_size 33554432
job_queue_processes 10
large_pool_size 16777216
log_buffer 1048576
open_cursors 300
optimizer_index_caching 90
optimizer_index_cost_adj 50
pga_aggregate_target 1073741824
processes 400
query_rewrite_enabled FALSE
session_cached_cursors 100
sga_max_size 1897893792
shared_pool_size 218103808
sort_area_size 524288
Thanks for your time and considerations.
Regards,
Shawon

Answer 1

Thank you very much Pierre Forstmann! Here is more information that u asked-
1. We already used statspack. But when the load avg went high- we did not run it.
2. We found the load avg using "top"
3. The users are complaining that the system is slow for few days. But suddenly it went over 20 and nobody can use the system.
4. In db server, we just run oracle and SharePlex. We have another host for web server. No other unix process are taking the CPUs, only oracle is taking it.
5. We can not produce the same load by running queries in sqlplus.
6. We did not try SQL tracing for one user connection and analyze with tkprof.
Someone of us ran a 20 minute sampling of system statistics when the system was pegged at a load average of 20, and then flushed the shared pool which forced the existing sql to get re-parsed on the next execution. The load average on the system started dropping within seconds of doing this, and then went all the way down to a level under 1 in a matter of minutes.
How did it help to improve the load avg? Since then the load avg is below 2. But we are not sure whether the problem is solved or not. Could you please tell me what is happening with oracle?
Thanking you.
Regards,
Shawon