Hardware configuration

[ranjitw]

Hi ,

Please let me know the minumium hardware configuration required for managing 1000 hosts with 10K services?
How much space need to be allocated for the database ?
Can anyone share standard architecture document ?

[jdalrymple]

There are too many variables for us to publish such a specification. I can offer some general guidelines.

I would say on average, plan on about 3MB of disk space per host/service monitored (database not included)
If you're network service heavy up that to 4MB per service
Regarding the database, it's typically pretty lightweight, 1MB per host/service should be plenty

Regarding memory and CPU it's very dependent upon what you're monitoring. If your 1000 hosts are ESX servers I'd say plan on 20 or more GB of memory and no less than 12 CPU cores. If your 1000 hosts are just Linux boxes running a handful of services you'd probably be fine with 8GB of memory and 2 CPU cores. Every type of monitored host/service has a different usage profile and as a result has different requirements from the monitoring system.

Guessing blindly - I'd say plan on a 6 core system with 12GB of memory and 80-100GB of disk space. That information comes free of any guarantees that it is actually fitting to your needs, it's a total guess based upon the information being made available to me.

[ranjitw]

Thanks for the valuable inputs.

Usually under which File system the database will be created ? Would 50 GB space be sufficient for database file system ?

Could you please let me know the FS and disk space to be located in general terms.

what mount points should be taken for backup incase of disaster recovery.

[jdalrymple]

We use default configurations in our install process. Bear in mind our install process places the database for you. By default it's under /var/lib/MySQL/. You can put it anywhere you want, just know that any customization you do makes it more difficult for us to troubleshoot in the event of a failure. Regarding "mountpoints best for DR" I don't know that I understand why any 1 mount point would be better than any other. The best option for DR (and performance) would be to offload the database to a dedicated host.

Regarding sizing, I already offered my (very general and potentially short-sighted) guideline of 1MB per host/service. Again - any guidelines I give you aren't reliable real world numbers. Because of the variance from one environment to the next we CAN NOT create a reliable sizing specification.

The best option is to virtualize everything so that you can later resize as needed.

[ranjitw]

Thanks , Please close the post.