Closed rtsisyk closed 8 years ago
alyapunov
commented
8 years ago I would add that QF implemented in vinyl is correct and I can explain its every line of code, QF itself is very cool data structure and could be useful somewhere else, but not in vinyl, at least not with used settings
kostja
commented
8 years ago @alyapunov investigated that quotient filter in vinyl doesn't work algorithmically with configured parameters. I checked coverage result and figured out that "optimal" case of this filter is never triggered even by vinyl/stress.test.lua. @alyapunov proposes to replace QF with a simple bloom filter. I believe that it is possible to enable bloomfilter for all indexes and stop confusing our users with unclear
amqfoption.@kostja Could we remove this option? There is no free resources to maintain dead code. One fool can write more code than four wise men can understand.
Go ahead.
Konstantin Osipov, Moscow, Russia, +7 903 626 22 32 http://tarantool.org - www.twitter.com/kostja_osipov
kostja
commented
8 years ago I would add that QF implemented in vinyl is correct and I can explain its every line of code, QF itself is very cool data structure and could be useful somewhere else, but not in vinyl, at least not with used settings
Move to salad then.
Konstantin Osipov, Moscow, Russia, +7 903 626 22 32 http://tarantool.org - www.twitter.com/kostja_osipov
@alyapunov investigated that quotient filter in vinyl doesn't work algorithmically with configured parameters. I checked coverage result and figured out that "optimal" case of this filter is never triggered even by vinyl/stress.test.lua. @alyapunov proposes to replace QF with a simple bloom filter. I believe that it is possible to enable bloomfilter for all indexes and stop confusing our users with unclear
amqfoption.@kostja Could we remove this option? There is no free resources to maintain dead code. One fool can write more code than four wise men can understand.
Bloomfilter or any other alternatives can be implemented after finishing with more important tasks.