I am new to clusters in general and expect them to be a bit slower than a stand alone system… but I am QUITE a bit slower…
Existing prod is a 4 node 32 gig system with the following settings
innodb_flush_log_at_timeout=1800
innodb_flush_method=O_DIRECT
innodb-log-files-in-group = 2
innodb-log-file-size = 1024M
innodb-file-per-table = 1
innodb-buffer-pool-size = 22G
innodb_stats_on_metadata=0
innodb_lock_wait_timeout=19000
innodb_io_capacity_max=1000
innodb-doublewrite=0
innodb-flush-log-at-trx_commit=2
innodb_log_buffer_size=4M
innodb_change_buffer_max_size=50
innodb_support_xa = 0
mysqlslap results in the following which is pretty good…
mysqlslap --user=xxxx --password=xxxx --host=localhost --create-schema=“test” --concurrency=30 --query=“insert into t1(name) values(‘111’)” --number-of-queries=100000 --iterations=1
mysqlslap: [Warning] Using a password on the command line interface can be insecure.
Benchmark
Average number of seconds to run all queries: 3.669 seconds
Minimum number of seconds to run all queries: 3.669 seconds
Maximum number of seconds to run all queries: 3.669 seconds
Number of clients running queries: 30
Average number of queries per client: 3333
The same on the cluster results in which is huge (using the same innodb settings and going to SSD storage… no spike on the systems while running the tests)
The clusters are on the same vlan with a 0.31 ms response time and 10 gig interfaces between them (inside the same enclosure) so network is not a issue
It does not change if I modify the slave threads The new clusters are 32 gig 16 core units
wsrep_slave_threads=6
wsrep_provider_options=“gcache.size=4G”
mysqlslap: [Warning] Using a password on the command line interface can be insecure.
Benchmark
Average number of seconds to run all queries: 112.825 seconds
Minimum number of seconds to run all queries: 97.198 seconds
Maximum number of seconds to run all queries: 124.199 seconds
Number of clients running queries: 30
Average number of queries per client: 3333
What am I missing?
**Note I know to run a 3 node cluster but just testing items right now