On Sat, Feb 12 2011, Santosh Rajan wrote: > My 2 cents before you jump into the nosql bandwagon. > > 1) If sql works for you, stick with it. RDBMS's like postgeSQL, MySQL > will not wake you up in the middle of the night with a crash. > 2) If you have scaling problems, add some horse power to you hardware, > battery backed RAID, and solid state hard drives are good for you. > (Prices have come down in the last year or two). > 3) Disk space is cheap. Avoid joins while using sql as far as > possible. Create additional table to do your indexing and grouping. > 4) And if you still think you need nosql, god help you.
[...] One of the things mentioned during the event was collecting logs from remote sites that have only access to the net for a short while every day. A data store like couch which works by appending documents is ideal for collecting log output. You keep dumping logs into it (over a local connection) and when you have access to the internet, you replicate all the local databases to a master couch database. I liked the idea and think it's an interesting way to approach the problem of synchronising logs. NoSQL databases are interesting and while RDBMs have their own applications, a lot of possiblities open up with document stores. To view them *purely* as alternatives to relational databases is, in my opinion, missing the point. >From your last point, I take it that you feel that one shouldn't even consider noSQL databases and somehow spend money and time squeezing performance out of relational databases. That reminds me of people who refuse to try out new languages and technologies and make engineering decisions and stick to, say, COBOL. -- _______________________________________________ BangPypers mailing list BangPypers@python.org http://mail.python.org/mailman/listinfo/bangpypers
