Multi threaded compaction helps there https://issues.apache.org/jira/browse/CASSANDRA-2191
It's another reason to avoid major / manual compactions which create a single big SSTable. Minor compactions keep things in buckets which means newer SSTable can be compacted needing to read the bigger older tables. It's also a reasonable factor to consider when sizing nodes. Cheers ----------------- Aaron Morton Freelance Cassandra Developer @aaronmorton http://www.thelastpickle.com On 15/08/2011, at 5:32 AM, Philippe wrote: > Hi, > > As on-disk SSTables become bigger and bigger because more data is added in > the ring, compactions take longer and longer because each file is becoming > bigger. > Isn't there a time where compacting will take so long that compaction just > can't keep up with the amount of data ? It looks to me like that's > independent from the write throughput, just a question of how long it takes. > > What am I missing ? > > Thanks > Philippe
