The birth of the open source enterprise stack
Posted Jun 29, 2006 6:01 UTC (Thu) by jwb
In reply to: The birth of the open source enterprise stack
Parent article: The birth of the open source enterprise stack
Well, I'd be happy to introduce you to some of these applications. For an example, I have written a system which caches the information about the last trade of hundreds of thousands of stocks. During active trading, this information comes at the rate of thousands of updates per second. The amount of information to be stored is basically fixed, as there are a fixed number of stocks to be traded. In MySQL, the database size stays nice and constant, I get atomic updates, and I also get a transaction rate up to 10k/sec on hardware costing less than $2000.
If I were to use PostgreSQL for this load, I would end up with a huge, ever-growing file, because updates are handled like inserts. I would have an always-running VACUUM process, but the file would probably grow slightly at peak times anyway. I would likely have unbounded index growth, as VACUUM does not reclaim index space (only REINDEX does this, and it requires exclusive access in some versions). In all likelihood, I would need a nightly maintenance period, which I can't afford because there are stock markets all over the planet.
So there's an example of a load where you'd better use MySQL.
to post comments)