Tuesday, November 18, 2014

My take on 2ed day 2014 Lucenerevolution at Washington, DC

Following is summary
Day started with SOLRCloud at Apple presentation by Apple team.  It is solid presentation in terms of  challenges faced during SOlrCloud implementation at Apple. Still they are implementing and seems to be they needed lots of new features (disaster& recovery space) and they are trying to automate as much as possible.  They gave Some JIRA tasks and they are contributing something back to community. Finally something back to community.  This session followed by lucid works presentation on scaling SOLR cloud for massive data. This story builds on top of Apple. Good one. After above general session, I attend following separate tracks.
a)     Solr on HDFS – Past, Present, and Future (Introductory) Running SOLR on HDFC & challenges.
Solid presentation by mark miller. 
This is just OK.  I felt it is pure hypothetical use case. He was saying some existing customer moved from Solr 3.1 to Solr 4.10 simply copying Solr 3.1 config & schema,xml files in to Solr 4.10 etc. etc.  I asked interesting questions after the session. Still lots of things are hidden. 
Solid POC, proof of concepts, effort.  (Basically with using freebase, he is trying to classify the content.)  At some point, I worked on similar POCs.  Mostly likely some site will use for some kind subscription based searches… I don’t think this make it big.
General talk on spatial search. Good one... Overall I am facing different problems in this space. For my employer geodistance is not much useful. I need a truce driving distance based search results. After the session also, I talked to David on this one. Seems to be this is not possible in the current SOLR Geo plugin space… I am thinking forking David’s original code and make it customer specific.  
Actually I know the challenges in Relevancy and it is difficult to cover in 30 minutes session.  Since not many alternatives and another friend dragged me to the session.  Routine talk. Search is not good enough. Relevancy is needed & you need to consider signal, end users, click throw, conversion rates … Don’t expect anything from Ph.D. guy in 30 minutes talk.
Simply superb.  Presentation started with problem statement. I.e. problem with image search & quick demo of how color search makes the difference. After demo he explained how they implemented with using SOLR.  Simply I loved the approach.

10 Keys to Solr’s Future Grant Ingersoll, Lucidworks  Typical Grant talk. Nothing special. I was not expecting too.

After Apple talk, I talked to the Apple  Sorl team, 2 folks, about the size of the search team. His response was Oh we are 8 to 10 people.  We are small & our productivity is great etc…  I stopped listening to him. He does not know he is talking to someone who delivered 3 large scale SOLR enterprise implementations single handly….

Overall an average 2ed day. Talked to few people on their SOlr challenges.
One of the interesting conversation is enterprise cloud search implementation by Hitachi folks.
Final thoughts… For 3 years in a row, I was attending lucenerevolution & now I am feeling most of the session are repeated. (May be I am doing too many things in SOLR.)
At this point, my take is I will not attend next year lucenerevolution. (90%, I decided on this one.)

I will enclose few picture later.




No comments: