Scaling Systems Insight Manager 5.x

[Jason Sylvester_1]

Are there any sizing guidelines for SIM as far as number of systems managed, number of current users, etc?

I have a single SIM server, also running VCRM, it's a ProLiant ML370 Win2003 SP1, SQL2000 SP4, dual 3.06GHz Xeon, 2GB RAM. Curently running SIM 5.0 SP2.

Assuming a wide area network (i.e. servers are scattered rather than local to the SIM server), what would be a reasonable number of systems that could be managed with this level of hardware? What hardware improvements (RAM, processor) would provide the most bang for the buck?

[Robert Romanczuk_1]

Jason,

There is actually a sizing utility that askes you a series of questions, then suggests a hardware configuration to run SIM on.

http://h71019.www7.hp.com/activeanswers/Secure/274366-0-0-0-121.html

I seem to recall also reading the suggested guidelines in some of the documentation, prehaps the installation guide. Unfortunately it escapes me at the moment. If you haven't been to the HP SIM home on www.hp.com though, you should check it out. It includes a link to the "information library" with some useful reading.

http://h18004.www1.hp.com/products/servers/management/hpsim/index.html

Hope that helps!

[Jason Sylvester_1]

That sizing tool was helpful.

But there are a couple limitations - every time I enter our current parameters (1100 nodes, Windows CMS) I get an answer "Sizing Failed" {heck, I already know that:p }

Also, the minimum speed that can be select for the question of the network between CMS and managed nodes is 10/100, which doesn't accomodate a realistic view of a WAN environment.

Is 1100 nodes too many for a single server in a global network? I know that my experience is slow and I am local to the CMS, the SIM users in remote locations have it even worse...

[David Claypool]

The best practice would be to locate the HP SIM server whereever the largest population of servers is. For status and event operations, the amount of data exchanged is relatively small, so those operations may not be too relevant.

Some best practices:

- SQL2000 is good for your population (>500 is the rule of thumb for SQL over MSDE)
- By default, data collection is run bi-weekly. This is the most data-intensive task as it collects configuration information for the database. You might consider breaking your servers into groups and running the collection at different times and intervals.
- If possible, disable the server status polling task that by default runs every 5 minutes. Try to create site lists and create new polling tasks for them. For sites with particularly poor response times, you may decrease the frequency of polling.

[Bill Slobodnik · Accepted solution]

I'm currently monitoring about 1800 servers, about 800 over a WAN using a Proliant DL380 G3 w/4gb ram. I am using sim for monitoring, email alerts and report generation, not for software updates or pushes of any kind.
The most significant step I took to improve performance was to migrate the SQL 2000 database to another server. When both SIM and SQL were on the same server, the data collection process would take about 12 hours and the CPU's would run between 90-100%. Since migrating SQL the DC process takes about 2.25 hours and the CPU rarely top 50% utilization.
For improvements, I would suggest to max out your ram and upgrade to HPSIM 5.0 SP3, I noticed a lot of the previous sluggishness of HPSIM improved after upgrading to SP3