Database Reference
In-Depth Information
create table tweets(tweet_id text primary key,body text)
with caching='rows_only' and bloom_filter_fp_chance=0.004
and populate_io_cache_on_flush='true'; // create table
alter table user with populate_io_cache_on_flush='true';
//alter table
It is recommended to keep this true for frequently queried column families which
can fit well within memory for better read throughput.
Figure 8-2
shows the table the
user created with this attribute set to true (see marked line).
Figure 8-2
.
User table enabled with populate cache on I/O operations set to true
Row Cache
With row level cache enabled, the entire row is available in cache. Row cache requires
a lot of space in comparison to key level cache. With Cassandra 1.2 onwards row cache
can be moved to off-heap, but deserialization of rows still happens temporarily in-heap.
Table 8-2
shows several row cache configurations and their descriptions.
Name
Description
By default it is disabled and set to 0.
row_cache_size_in_mb