Hi everybody, need some help with a perfomance issue.
Running 8.1B on a Sun E3000 with Solaris 2.5.1, which is properly patched (Solaris, not 8.1B thats not patched at all). 2 gigs ram and 2 336 MHz Ultra-II CPUs.
Upgrading Progress and Solaris right now is not an option.
We run about 200 local users and 90 remote (webspeed 2.1), normally about 30 of the webspeed users agents are busy, when loaded all remote users are busy. 99% Buffer hits and somwhere between 3 and 5000 reads/sec. Much more reads than writes. 3 apws, 1 biw. After imaging enabled.
When heavily loaded the system really slows down, being up to 5 times as slow as the normal condition.
Server startup params:
-L 25000 -n 400 -bibufs 25 -aibufs 37 -B 40000 -H zeus -S tdata97 -Mn 21 -Mi 7 -Ma 7 -minport 2701 -maxport 2750
Client startup params:
-db /db/apo9900/data/apo96t -e 500 -d ymd -l 500 -T /db/apo9900/tmp -t -p m1
The temporary catalogue (-T) is on the same disk as the database, but the disk doesn't seem to be loaded. We're not running raid, just Solstice Disksuite mirroring.
-spin is not used, and "Resource waits" in R&D/Other/Performance looks like this: (The server has been up for about 3 hours, and is not very heavily loaded right now)
Tot per min per sec per tx
Resource waits 2716 17 0.29 0.07
Are those values good or bad? Should we use spin, and in that case, what value?
Solaris seems to be running good, lots of idletime and not much diskio.
Could be interesting to know that when we added 2 more CPUS and 2 more gigs of ram (have a nice supplier who lend things to us) everything went alow SLOWER. Ever heard of that? (Eeproms had same revisions, 3.2.28).
Any suggestions?
Thanks you!
Running 8.1B on a Sun E3000 with Solaris 2.5.1, which is properly patched (Solaris, not 8.1B thats not patched at all). 2 gigs ram and 2 336 MHz Ultra-II CPUs.
Upgrading Progress and Solaris right now is not an option.
We run about 200 local users and 90 remote (webspeed 2.1), normally about 30 of the webspeed users agents are busy, when loaded all remote users are busy. 99% Buffer hits and somwhere between 3 and 5000 reads/sec. Much more reads than writes. 3 apws, 1 biw. After imaging enabled.
When heavily loaded the system really slows down, being up to 5 times as slow as the normal condition.
Server startup params:
-L 25000 -n 400 -bibufs 25 -aibufs 37 -B 40000 -H zeus -S tdata97 -Mn 21 -Mi 7 -Ma 7 -minport 2701 -maxport 2750
Client startup params:
-db /db/apo9900/data/apo96t -e 500 -d ymd -l 500 -T /db/apo9900/tmp -t -p m1
The temporary catalogue (-T) is on the same disk as the database, but the disk doesn't seem to be loaded. We're not running raid, just Solstice Disksuite mirroring.
-spin is not used, and "Resource waits" in R&D/Other/Performance looks like this: (The server has been up for about 3 hours, and is not very heavily loaded right now)
Tot per min per sec per tx
Resource waits 2716 17 0.29 0.07
Are those values good or bad? Should we use spin, and in that case, what value?
Solaris seems to be running good, lots of idletime and not much diskio.
Could be interesting to know that when we added 2 more CPUS and 2 more gigs of ram (have a nice supplier who lend things to us) everything went alow SLOWER. Ever heard of that? (Eeproms had same revisions, 3.2.28).
Any suggestions?
Thanks you!
