phoenix-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexander Batyrshin <0x62...@gmail.com>
Subject Optimisation for SELECT ... WHERE pk IN (pk1, pk2...,pkN)
Date Sat, 25 Jul 2020 13:20:02 GMT
 Hello all,
I have I quite big mutable table and need to select rows by primary keys.
The problem is that "SELECT … WHERE pk IN (pk1, pk2...,pkN)" (ROUND ROBIN POINT LOOKUP ON
$num KEYS OVER) do not use HBase bloom filter ‘ROW’.
This leads to performance degradation when more than one StoreFile per region exists (that
is near always true on mutable tables under heavy write)

Im wondering is there any way to optimisation?
Can we force-use ROW bloomfilter directly in Phoenix server side filters? Or is there any
way to patch HBase method StoreFileScanner.requestSeek() to use ‘ROW’ bloom?


Offtop: ROWCOL bloom is bad option, because it consumes too much spaces and doesn’t fit
into RegionServers BlockCache
Mime
View raw message