hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jim Kellerman <...@powerset.com>
Subject RE: HBase Bloom filters hash
Date Mon, 22 Oct 2007 20:18:19 GMT
> -----Original Message-----
> From: Andrzej Bialecki [mailto:ab@getopt.org]
> Sent: Monday, October 22, 2007 11:45 AM
> To: hadoop-dev@lucene.apache.org
> Subject: HBase Bloom filters hash
>
> Hi,
>
> I'm curious why the hashing function that these filters use
> is based on SHA-1 (which is relatively slow to compute)
> instead of a bunch of fast and simple non-cryptographic
> functions such as Jenkins' hash (see
> http://bretm.home.comcast.net/hash/7.html
> for the evaluation of Jenkins hash).

The reason for SHA-1 is that it was what came with the open
source bloom filter implementation we used. We've been focused
on just getting things to work and not on performance, yet.

If you'd like to open a Jira, it will be on the list of things
to do - sometime. If this is really important to you, how about
submitting a patch? There's mostly just the two of us, +
contributors, so we have to put our priorities on bug fixing
and making HBase robust before we get around to adding more
features or doing performance analysis.

We'd really like to get more contributions... the project would
mature much more rapidly.

Hope this helps.

---
Jim Kellerman, Senior Engineer; Powerset
jim@powerset.com

Mime
View raw message