Dear team,
In order to compare the efficiency of 'put' operation for RIAK where
storage_backend=riak_kv_eleveldb_backend with eleveldb itself I ran several
tests.
Tests were made under basho_bench.
Please comment the results – the average efficiency of RIAK is 4 times smaller
than average efficiency of eleveldb!
Data were placed into empty base. Entry configuration in basho_bench:
{key_generator, {int_to_bin,{uniform_int, 1000000000}}}.
{value_generator, {fixed_bin, 50}}.
Tests were made on one server:
8 Core Intel(R) Xeon(R) CPU E5450 @ 3.00GHz
16 Gb memory
elevelDB
basho_bench configuration:
--------------------------
{driver, basho_bench_driver_eldb_ps}. (driver
basho_bench_driver_eldb taken from installation package of eleveldb,
complemented by parameter write_buffer_size set to different instances of
eleveldb)
{mode = {rate, 2800}}. (number was taken to level out the
efficiency on load)
{duration = 10}.
{concurrent = 4} (as amount of vnode in RIAK)
{operations, [{put,10}]}.
eleveldb configuration:
--------------------------
write_buffer_size = 30-60 mb (according to documentation it’s
in RIAK)
max_open_files = 50
Results:
--------------------------
total = 14658677
req/sec min = 1081
req/sec max = 32390
req/sec average = 24431
RIAK
basho_bench configuration:
--------------------------
{driver, basho_bench_driver_riakc_pb}.
{mode, max}.
{duration, 10}.
{concurrent, 10}. (number was taken to get the highest
efficiency)
{operations, [{put,100}]}.
{riakc_pb_replies, 0}.
RIAK configuration:
--------------------------
version 1.2.0
{ring_creation_size, 4},
{default_bucket_props, [{n_val,1}]}
{storage_backend, riak_kv_eleveldb_backend}
eleveldb configuration: default RIAK settings
Results:
--------------------------
total = 3981611
req/sec min = 5214
req/sec max = 7645
req/sec average = 6636
Denis.
_______________________________________________
riak-users mailing list
riak-users@lists.basho.com
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com