incubator-cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sylvain Lebresne <sylv...@datastax.com>
Subject Re: any ways to have compaction use less disk space?
Date Wed, 26 Sep 2012 13:05:50 GMT
On Wed, Sep 26, 2012 at 2:35 AM, Rob Coli <rcoli@palominodb.com> wrote:
> 150,000 sstables seem highly unlikely to be performant. As a simple
> example of why, on the read path the bloom filter for every sstable
> must be consulted...

Unfortunately that's a bad example since that's not true.

Leveled compaction keeps sstables in level of non-overlapping key
ranges. Meaning that a read only has to check one sstable by level (a
little bit more to be precise since it has to include all of Level 0,
but provided you node is not lacking too much behind, that's still a
small amount of sstables). I'm too lazy to do the exact maths but I
believe that for 700gb you'll have 8 levels.

--
Sylvain

Mime
View raw message