Why MemSQL is slower than SQL Server for Select SQL with Substring Operations on Binary Columns -
i have table 2 binary columns used store strings 64 bytes long maximum , 2 integer columns. table has 10+ million rows , uses 2gb of memory out 7gb available memory, there plenty of available memory left. configured memsql based on http://docs.memsql.com/latest/setup/best_practices/.
for simple select sql binary columns compared values, memsql 3 times faster sql server, rule out issues such configuration or hardware memsql.
for complex sqls use
- substring operations in select clause ,
- substring , length operations in clause
memsql 10 times slower sql server. measured performance of these sqls on memsql taken after first few runs make sure sql compilation durations not included. looks memsql’s performance issue has how handles binary columns , substring , string length operations.
has seen similar performance issues memsql? if so, column types , sql operations?
has seen similar performance issues memsql substring , length operations on varchar columns?
thanks.
michael,
my recommendation: go rowstore, varbinary instead of binary, consider putting indexes on columns or creating persisted columns, , try rewriting predicate like.
if paste example query, can transform it.
the relevant docs here dev.mysql.com/doc/refman/5.0/en/pattern-matching.html docs.memsql.com/4.0/concepts/computed_columns
good luck.
Comments
Post a Comment