On Mon, Feb 22, 2010 at 2:23 PM, <jeremey.barr...@nokia.com> wrote: >> You could use supercolumns here too (where the supercolumn name is the >> thing type). If you always want to retrieve all things of type A at a >> time per user, then that is a more natural fit. (Otherwise, the lack >> of subcolumn indexing could be a performance gotcha for you: >> http://issues.apache.org/jira/browse/CASSANDRA-598). > > Would you say the supercolumn approach is faster than scanning rows? Any > particular advantages or disadvantages to writing to a bunch of supercolumns > at once (e.g. in one user row), vs. writing to a bunch of rows at once (with > the same key prefix, i.e. close together in an order-preserved store)?
Depends. What 598 says is that whenever you want to read any subcolumns from a supercolumn, it has to deserialize all the subcolumns. So if you were going to ask for all of them anyway, no big deal. But if you just want a relatively small number, it will be non-negligible. -Jonathan