lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "David Balmain" <>
Subject Re: Ferret's changes
Date Wed, 11 Oct 2006 01:42:43 GMT
On 10/11/06, Doug Cutting <> wrote:
> David Balmain wrote:
> > The start of my benchmarks are here:
> >
> >
> Ferret looks fast!  Nice work.
> A big knee in indexing performance occurs when indexes get much larger
> than memory, when merging requires a lot of disk i/o.  In these cases
> the algorithm can matter more than its implementation.  (The
> implementation only needs to be fast enough so that i/o dominates.)
> Would you expect Ferret to perform markedly differently than Lucene Java
> when updating, e.g., 50GB indexes on a machine with 512MB of RAM?

I don't like to say without testing. I have 90Gb of Gutenburg text
here so I'll give that a go a little later on. You are probably
correct though, the difference may be a lot smaller.

> Lots of indexes are smaller than memory and their updates don't require
> a lot of disk i/o.  It's important to be fast in that case too, and your
> benchmarks show Ferret doing quite well there.
> Doug

Thanks. That is especially the case in the Ruby community. Although I
would like Ferret working well for very large indexes, I don't see it
being used to build the next Google.


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message