What is the hardware requirement for OpenNMS systems that can manages 30K+ device? Is there any literature on hardware scalability of OpenNMS?
In those cases, more cores and spindles for random I/O will be handy. Most of my OpenNMS installations have been fairly small to medium (1000 devices). I can run those inside of dedicated VMWare virtual machines. I go with separate physical server for larger installations. An example:
1 or 2 Intel E5640 or greater CPUs.
6-8 10k or 15k SAS disks in a RAID 1+0 (One large logical disk, leaving the /opt/opennms partition as the growth partition). A SAN could also be handy.
And a lot of this depends on how much logging you enable (I disable debug logging once the install is stable) and how much data you wish to record about each node. If you're just looking at node-up/node-down alerts on 20K of those nodes, the demands will be far lower than full SNMP polling and graphing of each. There are also a few optimization tips on the Postgres database side, as well as an option to leverage ramdisks for the RRD files.
What types of nodes are these? Mostly networking or servers or workstations?