An interesting way to pass the time is the 'Random Article' feature on wikipedia. You press it and get one of 2.2 million random pages. Often they are about obscure American towns but today it turned up Amazon SimpleDB. This looks interesting, an online database where you pay for the amount of data you use.
From what I can make out, data is stored in a huge hash, although each hash entry can hold multiple items (a bag in a hash?). They compare it to a worksheet but that analogy seems to be aimed at pointy haired bosses who can't think beyond Microsoft Office. The api looks very simple, it shuns SQL.
Advantages:
- someone else worries about scalability, administration, backups
Disadvantages:
- not SQL: not a bad thing but it could give problems when used with applications written for SQL (i.e. most of them).
- inversely, if you developed an application that relied on this, you better hope they keep it running.
So, it's interesting but would I want to rely on it? Probably only if it had an SQL shim on top so I could migrate away pronto, should the need arise.
Google for SimpleDb and there is lots of analysis around. 1024 bytes per item? Lexicographical comparison of dates and numbers?
I think I'll stick with SQL.
Peter's Moblog
