Massive deployment, expected problems?

[tweidner6471]

So, I'm in the process of building out a rather large vicidial installation. I'm currently only building out to about 200 but eventually plan to scale to the 1000-1500 range.

Out of curiosity, has anyone ever experienced any issues with the system that aren't typically associated with asterisk (meaning weird deadlock issues, etc)?

I have a very good estimate of what equipment is needed, just want to know if anyone has any good advice or any 'gotchas' that I might encounter.

I do plan on trying to pull Matt aside once this starts to take place, but I figure until then it doesn't hurt to ask the community at large about the things they've seen an experienced.

[mflorell]

Buy the best DB server you can find. The most RAM, CPUs, fast harddrives and a hardware RAID you can get. Everything else scales by adding more machines.

[tweidner6471]

Currently the DB server is built with the following specs:

RAID50 80GB SSD's (gets around 1700 IOPS)
Dual Quad Core 3GHZ
32GB RAM

We're going to build a few of these out for transactional replication and try to get some read/write offloading onto some slave systems to help with performance.

As a side note, have you had any positive/negative experiences with utlizing any HTTP load balancing setups, such as LVS?

[mflorell]

I wouldn't recommend any RAID with a "5" in it. We go RAID 10 most of the time.

I would also recommend going for a machine with more CPUs and more RAM before attempting to break off into a multi-server database solution.