After our two successful hackdays, it was on to the main event of the week and the largest open source search event of the year. In between catching up with other Lucene/Solr folks on the first day I enjoyed Chris ‘Hossman’ Hostetter’s talk on Hidden Gems of Apache Solr with some great tips on obscure Solr query syntax, and Bloomreach’s fast-paced talk on the SolrCloud Rebalance API which allows one to autoscale large Solr systems (although this feature isn’t quite yet available in Solr 6, we’re promised it’s being worked on). I then had a pleasant partner lunch with Lucidworks and heard about some exciting developments for their Fusion search platform – we can expect to see version 3 of this Solr-based product soon and I’ll be blogging more about this in coming months. Dragan Milosevic‘s talk on how aggregration performance compares between various systems including Elasticsearch and Solr was slightly hamstrung by his laptop failing, but he bravely carried on and led us to the (unsurprising) conclusion that both perform pretty well but working with HBase directly can be faster due to its single index. The day finished with a party held on the 50th floor of a nearby building, with fantastic views over the city.
On the second day I caught Scott Blum‘s talk on another autoscaling strategy for Solr. They have a 13 billion document index running on 6000 Solr cores on 32 nodes so have had issues with JVM garbage collection which they initially solved with a manual reload of each core every few hours. They eventually built Solrman which can automatically balance Solr cores across a set of nodes – this is a solid alternative to the Rebalance API and we’ll be looking at both for some of our clients soon. We followed Scott with our talk on Coffee, Danish and Search (Powerpoint slides are here and video here) about our work on a multilingual media monitoring system for Infomedia which was well received. Having stepped out for a minute I was unable to get back in the room for Kevin Watters‘ talk on the Solr Graph Query which was extremely popular! Grant Ingersoll finished the conference with a closing keynote and we then headed for the airport.
Thanks to Lucidworks and the conference sponsors for another great event – it felt busier than previous years and this is more evidence of the ongoing healthy state of the Lucene/Solr community. We’ll be back!
