SAM server configuration question.

We bought two new HP dual Xeon quad-core CPU servers.

The SQL server has 4 15,000 RPM hard drives and is set up in a RAID 10. There is no partition, and the OS and SQL database will reside on the same physical volume (s) in the RAID array.
The SAM server has two hard drives and will be setup in a mirrored configuration (RAID 1) for redundancy.
Both servers have dual NICs, and are using HP network teaming to load balance
We use SAM, NPM, NCM, UDT, and Network Atlas

The question was brought concerning the SQL server on having the OS and SQL database on separate partitions. The reason was keeping them on separate logical volumes would help reduce unnecessary disk queue times brought on by the OS (which may affect the SQL server writing). Right now we are having issues with slow response times when drilling down on nodes, but I honestly feel our current SQL database server is taking on too much (it only uses one hard drive for the OS and SQL). I feel it is also over burdened based on the hardware capability vs. number of polling.

We only have 4 drives so taking one out to put the OS on will leave us with only 3 drives for the SQL database (which means no RAID 10). I'm thinking with the speed of the 4 drives in a RAID 10, 16GB of memory, and a dual CPU quad-core Xeon setup ....that having the OS and SQL database reside in the same array won't make a difference speed wise.

Any thoughts or experiences with this ? We don't have the cash or drive bays to create multiple RAID arrays. We've gone the SAM server with 2 SAS disk drives, and the database server with 4 SAS disk drives.

Find more posts tagged with

server_details

solarwinds_sever_&_application_monitor

raid_10

server_load

sam_5.0

Accepted answers

All comments

aLTeReGo

How does the CPU look on the application server? Is it normally high? If so, you may benefit from moving the web interface off the poller onto it's own virtual or physical machine. There's nothing of much value I could suggest beyond this that would improve performance dramatically without altering the array configuration. It's unusual that we encounter a system that's suffering from performance issues that's just running SAM, NPM, UDT and NCM. It may be that the Orion server is receiving a large volume of syslog or trap messages. What is the current size of your database and what are your used license counts for each module? Also, have you modified your retention settings?

One thing I would recommend is to work with your DBA to create a maintenance schedule for the database that rebuilds the indexes nightly and defragments the tables. This should result in a noticeable boost of performance if you don't already have a maintenance plan defined on the database.

gregatkins

We run traps and syslogs on a huge amount of switches. Our database size now is 57GB. I work at a casino, so have a huge amount of syslogs and traps in case of security issues or trying to identify possible problems.

aLTeReGo

You may want to consider looking at our Log & Event Manager which has been purpose built to handle a steady stream of syslog and traps. Even moving this data off onto Kiwi Syslog should help improve performance. Other alternatives would be to limit/reduce your data retention periods for syslog and trap data or follow the steps outlined above for reducing table and index fragmentation.

gregatkins

We've considered the log and event manager as a possible solution. I ran the database maintenance tool, plus ran the SQL query to defragment the tables. Right now our SNMP traps are set for 180 days, which I think is excessive. I'm going to suggest keeping them for 60 days. I work at a casino, so there are always rules and policies on data retention. . There are also some minor "discussions" going on because people expect the new server to load the UDT summary page instantaneously, when that happens to be the largest table in the database.

I'm also using the old Microsoft tool SQL I/O Sim , that is pretty helpful on doing SQL stress tests to tweak the performance.

I should have Solarwinds tech support on a teleconference call during one of our weekly IT meetings