Server Performance after RAM upgrade
Pervasive SQL 9.1
Windows 2000 SP 4 on Intel XEON 2ghz machine
I was having performance issues on this server on one particular query. I copied the entire database to a newer machine and had no problems. The query would run for up to 60 seconds on the older machine, in my web application it would of course time out. On the newer machine the query ran anywhere form instantly to 2 seconds.
The old machine only had 500 MB RAM so my first choice was to upgrade the RAM. So I added 2 GB to the old server.
Here is the problem. Now all of the queries run in about the same amount of time. 7-12 seconds. That is of course a big improvement on the one query I was having trouble with and that speed is almost acceptable. The bad thing is that the queries that ran instantly on the old RAM configuration now take as long as the slow one (7-12 seconds)
I monitored the database while accessing the website and the lag time is in the database or at least it is staying connected to the table while the wait time is happening.
My Processor load never goes over 57% during the queries. Some of these (now) slow queries are single row indexed selections
I have attached the configuration for performance tuning and memory usage. I have not adjusted them.
config1.gif
config2.gif
Windows 2000 SP 4 on Intel XEON 2ghz machine
I was having performance issues on this server on one particular query. I copied the entire database to a newer machine and had no problems. The query would run for up to 60 seconds on the older machine, in my web application it would of course time out. On the newer machine the query ran anywhere form instantly to 2 seconds.
The old machine only had 500 MB RAM so my first choice was to upgrade the RAM. So I added 2 GB to the old server.
Here is the problem. Now all of the queries run in about the same amount of time. 7-12 seconds. That is of course a big improvement on the one query I was having trouble with and that speed is almost acceptable. The bad thing is that the queries that ran instantly on the old RAM configuration now take as long as the slow one (7-12 seconds)
I monitored the database while accessing the website and the lag time is in the database or at least it is staying connected to the table while the wait time is happening.
My Processor load never goes over 57% during the queries. Some of these (now) slow queries are single row indexed selections
I have attached the configuration for performance tuning and memory usage. I have not adjusted them.
config1.gif
config2.gif
ASKER
I tested my query from control center and it runs instantly.
I watched the active user monitor and it shows overwhelmingly Cache access for all users. A ratio of about 2000 cache accesses for every 1 disk access.
We were getting ready to do the 9.5 patch when we noticed the slowness after ram.
I'll change the L1 cache next chance we get to restart.
Also the windows swap file was set at 4gb minimum and maximum.
I watched the active user monitor and it shows overwhelmingly Cache access for all users. A ratio of about 2000 cache accesses for every 1 disk access.
We were getting ready to do the 9.5 patch when we noticed the slowness after ram.
I'll change the L1 cache next chance we get to restart.
Also the windows swap file was set at 4gb minimum and maximum.
If the query runs instantly, then the query or the database is not the problem. Perhaps you have a problem in the web server with name resolution to the database. Make sure that the DNS servers are set correctly on the server, and that it you do a DNS lookup, you get the right address instantly.
You could also try getting a network capture from the web server doing a single function call, and see where the time goes...
You could also try getting a network capture from the web server doing a single function call, and see where the time goes...
ASKER
Okay, it is sporadic.
The same query I ran that was instant on the local Control Center a few munites later was taking 10 seconds or more. And the sporadic nature is the same on the local control center and the web site. The local network users are also seeing sporadic slowness.
Since this happened right after a RAM upgrade does it sound like it could be bad RAM?
Can I set my L1 cache to 1 GB? I currently have it set at 100 MB and my available RAM in Task Manager shows 1.8 GB avail
The same query I ran that was instant on the local Control Center a few munites later was taking 10 seconds or more. And the sporadic nature is the same on the local control center and the web site. The local network users are also seeing sporadic slowness.
Since this happened right after a RAM upgrade does it sound like it could be bad RAM?
Can I set my L1 cache to 1 GB? I currently have it set at 100 MB and my available RAM in Task Manager shows 1.8 GB avail
ASKER CERTIFIED SOLUTION
membership
Create a free account to see this answer
Signing up is free and takes 30 seconds. No credit card required.
ASKER
Well now I can't get it to slow down....
ASKER
I will come back after I get a chance to upgrade to 9.5x and reset the memory settings on the server. This machine is in another city and can't be restarted during the day. I have to schedule the down time with the client.
Thanks for your help.
Thanks for your help.
ASKER
I upgraded to 9.5x on the server engine and all of the clients. Oddly enough after the first morning of the RAM upgrade I haven't had the slow down again. I am accepting this answer because of the memory performance tuning suggestions.
THanks
THanks
1) Increase the L1 database cache from 100MB to 400MB. With 2GB of RAM, you have plenty of room for L1 cache, which is where the real work gets done, anyway.
2) Try patching the engine. V9.1 can be a bit on the unstable side, and num erous improvements were added in both v9.50 and in v9.52. Get both from Pervasive's web site. You must apply v9.50 patch FIRST, and then do the 9.52 update afterwards. Be sure to do both clients AND the server.
3) Monitor the database using Pervasive's Monitor and watch the database connections from the Active Users screen. See how many disk versus cache accesses are being generated.
4) Try running your queries manually from the Pervasive Control Center (PCC) and see if you see the same results there. It could be a connection-delay issue and not related to query performance at all.
5) If they are still slow there, try enabling the Query Plan Viewer and review the query to make sure that indices are available for the query to optimize on.