Wanting to scale up Cinemon's SNMP performance. We're currently in the 10-20,000 device range, which is okay, but I'd like to increase that by a factor of 5 or 10 to address some performance issues. We've got plenty of memory for that scale (though we'll need to work on the long-term storage system a bit). The big limiter right now is processor load in encoding/decoding all the SNMP in Python (and the long-term database (postgresql) storage).
Thinking I'll write a NetSNMP wrapper and wrap that into TwistedSNMP so that it can use either back-end. That should let me test the performance of the back-ends. I'm assuming NetSNMP can handle scanning 100,000+ networks without breaking a sweat on the server. It can probably handle 250,000 too, after all, 15x faster than Python isn't all that hard for C, particularly when it's all grotty, low-level stuff like SNMP parsing.
Not going to get done this week, though. Two deadlines on Friday to deal with.
Pingbacks are closed.