From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <vdavydov.dev@gmail.com>
Date: Wed, 28 Mar 2018 20:19:30 +0300
From: Vladimir Davydov <vdavydov.dev@gmail.com>
Subject: Re: [commits] [tarantool] 03/04: bloom: optimize tuple bloom filter
 size
Message-ID: <20180328171930.7n37ebb4ihvh5hjx@esperanza>
References: <20180327210817.GB11829@atlas>
 <20180327210938.GC11829@atlas>
 <20180328111615.5fltx74ujvp26dry@esperanza>
 <20180328170348.GK11829@atlas>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20180328170348.GK11829@atlas>
To: Konstantin Osipov <kostja@tarantool.org>, tarantool-patches@freelists.org, commits@tarantool.org
List-ID: <tarantool-patches.dev.tarantool.org>

On Wed, Mar 28, 2018 at 08:03:48PM +0300, Konstantin Osipov wrote:
> * Vladimir Davydov <vdavydov.dev@gmail.com> [18/03/28 14:50]:
> > > 
> > > In future, It would be nice to actually have stats about actual false
> > > positive rates so that we know if the papers are BS or not.
> > 
> > We already have bloom.hit and bloom.miss stat counters in index.info()
> > so we can calculate the false positive rate as
> > 
> >   bloom.miss / (bloom.hit + bloom.miss)
> 
> Do you have a test verifying that the formula works?

vinyl/bloom.test.lua uses these stats to check that observed false
positive rate doesn't exceed the configured value.