Can you just maintain the site-wide top 5 by combining top 5 of each Shards?
It can be done in memory and just O(1) operations. No Lucene is needed.
--
--
Chris Lu
-------------------------
Instant Scalable Full-Text Search On Any Database/Application
site: http://www.dbsight.net
demo: http://search.dbsight.com
Lucene Database Search in 3 minutes:
http://wiki.dbsight.com/index.php?title=Create_Lucene_Database_Search_in_3_minutes
On 2011/3/4 9:59, BrightMinds Dev wrote:
We are developing a large 4-tier multi-server app that will accept
Questions and related Comments supplied by users. There will be
100K's of users that live in Shards. Also, ideally there would be no
delay in adding content and seeing it in recent results but to make
the system performant a delay is acceptable.
On the main page we will have 2 panels with the 5 most recent
site-wide Questions and 5 most recent site-wide Comments.
On the user's profile page we would display similar panels except they
would only consist of links that pertain to that user.
There are essentially 4 design choices:
1) Do periodic DB SELECT calls and cache for site-wide content "AND"
do live DB SELECT calls for user specific content. Not wonderful but ok.
2) Use DB triggers to manage tables for site-wide content, though as
content is sharded we would need to aggregate and resort based on
results obtained from all shards "AND" do live DB SELECT calls for
user specific content. Sounds awful... .
3) Use a solution like memcached however having multiple servers add
and prune site-wide content to cache seems like it would be a synchro
nightmare "AND" do live DB SELECT calls for user specific content.
Nix that... .
4) Use Lucene to obtain most recent site-wide content (and better yet
push it into memcached) "AND" use Lucene to retrieve live user
specific content
Is this a natural fit for Lucene?
I understood Lucene is very performant but is finding the last X
documents based on a long (timestamp) something that it would do very
well?
Or is Lucene not appropriate and should we be considering something else?
Personally, I think Lucene is the best choice... but would like to
hear the thoughts of others.
Thanks,
--Nikolaos
---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscr...@lucene.apache.org
For additional commands, e-mail: java-user-h...@lucene.apache.org
