jackrabbit-oak-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jukka Zitting <jukka.zitt...@gmail.com>
Subject Re: Large-scale read benchmarks
Date Wed, 13 Mar 2013 11:27:02 GMT
Hi,

On Tue, Mar 12, 2013 at 6:26 PM, Jukka Zitting <jukka.zitting@gmail.com> wrote:
> Here are initial numbers from these benchmarks, executed against the
> SegmentMK on a m1.medium EC2 instance with a local MongoDB backend:

A few more results with different segment cache sizes:

    http://people.apache.org/~jukka/2013/LinearReadEmpty.png
    http://people.apache.org/~jukka/2013/UniformReadEmpty.png

The graphs show the 50th and 90th percentiles of the time it takes to
access 1000 empty documents. The horizontal axis indicates the segment
cache size that was used, and the vertical axis the time in
milliseconds it takes to access the documents.

For the linear reads it looks like the time it takes to fetch a
segment from the underlying repository is insignificant compared to
the time it takes to traverse documents in it, since the segment cache
size doesn't seem to have a noticeable impact on performance.

For random, uniformly distributed reads the cache plays a major role.
As a bigger percentage of the working set fits inside the cache, the
faster such random accesses become.

BR,

Jukka Zitting

Mime
View raw message