Hi,

I have recently been using the TSdbi/MySQL library in R, which allows mapping between time series matrices and data stored in a mysql backend - its actually pretty good, I've made loads of progress in a few days...

However I now now looking at scaling it up to many parallel sources of time series data, and my limited understanding of implementing this type of schema suggests that column orientated databases like monetDB should perform better than MySQL for this type of data. (?)

The schema suggested as the default for storing irregular time series is as follows; create table T ( id VARCHAR(40), date DATETIME, v double DEFAULT );

with an index on the ident and the DATETIME field; CREATE INDEX Tindex_id ON T (id); CREATE INDEX Tindex_date ON T (date);

So I am looking to support 500+ writes per second, while entertaining a whole bunch of queries of time periods like; SELECT date, v FROM T WHERE id = 'series1' order by date select date, v from T WHERE date > NOW() - INTERVAL 1 WEEK AND date < NOW();

I was hoping for some advice on whether monetDB is suitable for this type of data, and performance suggestions to get it blisteringly fast!!

Thanks,

Tom