I have a multiple field structure in which i have 2-3 fields with text that
will be searchable and besides that i have 3 very write intensive fields
(counters) in this structure. Those fields will only need to be visible in
results but will not be a part of the search query. How to update those
counters in the most efficient way ? So without reindexing anything etc.
How to handle them best? What kind of performance (updates per second) we
are talking about on updating those counters lets say on one Xeon E3-1230
v3 or any other cpu?
Performance is based on your dataset, you'd have to test it yourself.
Ideally you want to separate hot and cold data, so that when you do update
you don't need to reindex the whole document (which is what an update does
in ES).
You can do this via parent/child relationships, or having two indices with
each part of the doc that you then join them externally, as two examples.
I have a multiple field structure in which i have 2-3 fields with text
that will be searchable and besides that i have 3 very write intensive
fields (counters) in this structure. Those fields will only need to be
visible in results but will not be a part of the search query. How to
update those counters in the most efficient way ? So without reindexing
anything etc. How to handle them best? What kind of performance (updates
per second) we are talking about on updating those counters lets say on one
Xeon E3-1230 v3 or any other cpu?
Thanks for replying.
So basically by making those 3 int counters inside child i will only
reindex those 3 ints yes ?
Dataset for now is around 50k documents.
Then what's the performance of only reindexing 2-3 ints on lets say
performance of Xeon E3-1230 v3 or any E5 or i7 etc or any other cpu. I
would like an estimate, just to know if we are talking about ~50k updates
per sec or ~10k or ~1k depending on cpu. Just an estimate would be really
helpful.
On Saturday, March 28, 2015 at 6:50:45 AM UTC+1, Mark Walkom wrote:
Performance is based on your dataset, you'd have to test it yourself.
Ideally you want to separate hot and cold data, so that when you do update
you don't need to reindex the whole document (which is what an update does
in ES).
You can do this via parent/child relationships, or having two indices with
each part of the doc that you then join them externally, as two examples.
On 28 March 2015 at 16:27, Lorne Sorn <dulo...@gmail.com <javascript:>>
wrote:
Hello,
I have a multiple field structure in which i have 2-3 fields with text
that will be searchable and besides that i have 3 very write intensive
fields (counters) in this structure. Those fields will only need to be
visible in results but will not be a part of the search query. How to
update those counters in the most efficient way ? So without reindexing
anything etc. How to handle them best? What kind of performance (updates
per second) we are talking about on updating those counters lets say on one
Xeon E3-1230 v3 or any other cpu?
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.