We have a cluster with 10 server, replication-factor 2, using Cassandra 0.8.7., and . we are absolute newbies with cassandra ;-)
In column family "A" we are writing 40.000 rows per second for 72 hours (no reads, only reads on column family "B"). After that we insert for one hour 10.000.000 rows to column family "B". Then "A" is truncated and the inserts to column family "A" starts again. Additional there is a column family "C" which is "organizing" all that, with 5.000 writes and removes per second (only 1 - 5.000 rows at one time, but a lot of garbage). What are the optimal settings for these 3 column families? Should we start any maintenance jobs at specific moments ? For example a "compaction" on "B" after the 1 hour insert, or a "delete snapshot" after the truncate , or is cassandra doing everything optimal on her own ? With regards, Jannik
