Database Reference
In-Depth Information
Write pattern
When writing data, you can specify how it should be distributed across the rows and
clusters. There are four things that may be specified as follows:
•
Partitions
: This attribute specifies the distribution of the
INSERT
queries in each
batch across all partitions (clusters). The default is
FIXED(1)
, which means one
value per partition per batch.
•
Pervisit
: This shows the ratio of rows that goes into a partition; this ratio is propor-
tional to the total number of rows for the partition. If you have a secondary key
column distribution as
FIXED(20000)
, then
FIXED(1)/2000
will denote that
each batch will insert 10 rows. If not specified, it defaults to
FIXED(1)/1
.
•
Perbatch
: This shows the ratio of rows each partition should update in a single
batch, as a proportion of the number of rows picked by the pervisit ratio and parti-
tion count. The default is
FIXED(1)
.
•
Batchtype
: By default, the stress test uses the
LOGGED
batches. One may set it as
UNLOGGED
for better performance.