Another Database Post - Sorry

Question

I'm sorry to ask this question to everyone and I've already been through just about every previous posting about database questions.

My question is this:

I am trying to suggest specification to our DB guys but I just don't know the  answers and the documentation from solarwinds on the subject of database sizing etc is all but non existent.

I will be monitoring approximately 150 nodes and maybe up to 2000 elements or more, I don't know the frequency as yet etc.

I will be using cirrus on the nodes (50 ish) and also the netflow module, application monitor and syslog.

Our server team are suggesting that we might not need a separate installaton of sql but I think that we do.

I just can't back anything up with suggested sizing etc.

What size should by sql databases be (one for netperfmon and one cirrus i think will be created)?

I know it's a bit of a piece of string so just some ideas would be great.

Some people talk of having their database up over 150gig?

I'm also not sure about whether to use the HSE on a virtual server or a physical box. We are not buying all the modules for the standby box so maybe just using a virtual server for the primary is sufficient for resilience as the vmware should do it's own failover.

Sam

menz · Answer

Thanks for both of your advice.

I wish that Solarwinds provided more indepth documentation on the database part of the installation.

All your info is greatly appreciated.

Sam

aLTeReGo · Answer

I have a very similar configuration to yours and my database is around 100GB with one week of netflow data being stored in the database. I'm not using a separate SQL server for Orion, but instead I'm using a shared SQL server that is dedicated to doing nothing but SQL. In other words my pollers etc are on a different box than my SQL server, but my SQL server services other light databases in addition to Orion.

Network_Guru is right though. NTA is a monster when it comes to disk space. Whatever you think you might need triple it. Disk speed & I/O performance is the most important component to NTA. I recommend the fastest disk technology possible. I'm using 15k SAS drives. RAID 0 is an excellent choice for performance but if you need redundancy raid 10 is the only way to go.

When it comes to disks, the more the better. Though capacity is great, performance is determined by the number of spindles (disks) in the array. As it is, you'll need a minimum of four disks for a RAID 10, but eight to ten disks is ideal if you plan to run NTA.

Remember that disks loose 20% of their performance when they reach 80% capacity.

If you're going to share a SQL Server with other databases make sure the Orion DB is on it's own disks, or it could have a negative impact on the performance of your other databases.

The last piece of advice I could give you is to break out the database and transaction logs onto separate disks. My transaction logs are on a RAID1 and my Orion DB is on a RAID 10.

Network_Guru · Answer

With 2000 elements you should be OK running the Orion DB on the same server as the polling engine & web server.
With aggressive polling your Orion DB will be around 10 - 15Gb.

The Netflow DB should go on a dedicated SQL server - it is a very DB intensive app & can consume gobs of disk space.
I'm using a competitors Netflow product and currently use 250Gb of 500Gb Raid 0 disk array polling about 50 interfaces.