That sounds silly imho, if I'd write an app using lucene I'd always
implement some basic caching to speed up a number of things that would
just be plain stupid to query everytime.
You also have to keep in mind that ES is optimized for a distributed
setup so it adds some overhead.
Under the covers it is "just" Lucene with a lot of smartness added to
it. I think it would be very difficult to fairly compare the two.
Jaap Taal
[ Q42 BV | tel 070 44523 42 | direct 070 44523 65 | http://q42.nl |
Waldorpstraat 17F, Den Haag | Vijzelstraat 72 unit 4.23, Amsterdam |
KvK 30164662 ]
I wonder what kind of performance you want to compare. ES is based on
Lucene. Although ES implements its own caching, there is no much sense
in "stopping" caching because many functions rely on it. A field cache
is not a turbo charger, in an inverted index, it's simply essential to
cache fields. Note, Lucene brings its own field caching.
Jörg
Am 21.03.13 16:28, schrieb Ophir Michaeli:
How do I stop all caching?
I need to stop it for performance tests vs lucene.
While some say this is silly, I'd love to hear what you find - what you get
when you run ES vs. Lucene test where you have all ES-only cached disabled.
Please share your findings.
you should be able to set the filter.max_size = 0 per segment via:
index.cache.filter.max_size
see also:
further you can simply set "_cache" = false on each filter which should
prevent the filter from being cached.
You can also prevent the field_cache from being reused but I don't think
this makes any sense since you don't wanna benchmark uninverting which is
the same thing in plain lucene or ES.
If you only execute queries then you don't need to do anything since ES
doesn't cache queries nor their results.
Hope that helps,
simon
On Sunday, March 24, 2013 10:29:28 AM UTC+1, Ophir Michaeli wrote:
We have our own distributed system that uses lucene, and we're considering
changing to elasticsearch.
Our system has billions of indexed documents and the first tests are with
tens of millions,
so we want to check that the elasticsearch results are not influenced by
cache which might function different with billions of documents.
On Thursday, March 21, 2013 5:28:34 PM UTC+2, Ophir Michaeli wrote:
Hi,
How do I stop all caching?
I need to stop it for performance tests vs lucene.
I tried this at the yml fiel, but not sure that it covers all:
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.