You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-user@lucene.apache.org by Rajiv2 <ra...@gmail.com> on 2009/02/10 16:13:46 UTC
Feedback needed on sharding/distributed search
Hello, we’re currently in the midst of re-designing our search hardware
architecture and I have some questions about sharding and distributed
search.
1. What is the benefit of using sharding/distributed search over keeping the
index intact?
2. What is the best approach to determining shard size? Does it depend on
CPUs available or does it depend on other factors like expected traffic,
query response times etc. or both?
Currently we have two UI servers w/ the full index serving requests. Our
index will be getting much bigger and we’re going to have 2 to 3 Solr
servers. Each will have several distinct shards on them, using the
multi-core functionality, to serve requests. So if we decide to split the
index into 3 shards each server will have all 3 distinct shards.
Regards,
Rajiv
--
View this message in context: http://www.nabble.com/Feedback-needed-on-sharding-distributed-search-tp21935743p21935743.html
Sent from the Solr - User mailing list archive at Nabble.com.
Re: Feedback needed on sharding/distributed search
Posted by Yonik Seeley <ys...@gmail.com>.
On Tue, Feb 10, 2009 at 10:13 AM, Rajiv2 <ra...@gmail.com> wrote:
> 1. What is the benefit of using sharding/distributed search over keeping the
> index intact?
Primarily response time of single requests. If your response times
are fast enough with a single index, then simply replicate that index
for fault tolerance and scalability with increasing load (# of users).
The primary use case for distributed search is when the index is too
big for a single request to execute in a reasonable amount of time.
-Yonik
> 2. What is the best approach to determining shard size? Does it depend on
> CPUs available or does it depend on other factors like expected traffic,
> query response times etc. or both?
>
> Currently we have two UI servers w/ the full index serving requests. Our
> index will be getting much bigger and we're going to have 2 to 3 Solr
> servers. Each will have several distinct shards on them, using the
> multi-core functionality, to serve requests. So if we decide to split the
> index into 3 shards each server will have all 3 distinct shards.
>
> Regards,
> Rajiv
>
> --
> View this message in context: http://www.nabble.com/Feedback-needed-on-sharding-distributed-search-tp21935743p21935743.html
> Sent from the Solr - User mailing list archive at Nabble.com.
>
>