Why is my client CPU utilization so high when I use a cassandra cluster? -
this follow-on question why cassandra throughput not improving when add nodes?. have configured client , nodes closely recommended here: http://docs.datastax.com/en/cassandra/2.1/cassandra/install/installrecommendsettings.html. whole setup not world class (the client on laptop 32g of ram , modern'ish processor, example). more interested in developing intuition cassandra infrastructure @ point.
i notice if shut down 1 of nodes in cluster , run test client against it, throughput of ~120-140 inserts/s , cpu utilization of ~30-40%. when crank 6 nodes , run 1 client against them, see throughput of ~110-120 inserts/s , cpu utilization gets between ~80-100%.
all tests run clean db (i delete db files , restart scratch) , insert 30m rows.
my test client multi-threaded , each thread writes exclusively 1 partition using unlogged batches, recommend various sources schema mine (e.g. https://lostechies.com/ryansvihla/2014/08/28/cassandra-batch-loading-without-the-batch-keyword/).
is cpu spike expected behavior?
Comments
Post a Comment