This week we are attending elasticon 2016. If you want to follow us along you can check our twitter feeds. If you like to read the summary read this and the previous posts. Our twitter handles are:
All About Elasticsearch Algorithms and Data Structures
(Zachary Tong and Colin Goodheart-Smithe)
This was the most complex talk we attended this year. To be honest, it was a bit to complex. So if you are interested in this topic, wait for the video to appear and look at it yourself. The only few remarks I have are about filteres and their caching mechanism.
Filters work on immutable segments, therefore we can keep them for a long time. There are three different mechanisms that all have their own advantages mostly based on the amount of documents in the segment. Luckily the choice for the mechanism to use is chosen automatically.
So depending on the amount of values in a segment different ways to store the filter results are used:
- Less than 4096 documents, a Sorted List is used,
- More than 4096 documents, a Bit map is used,
- A lot more documents, a Compressed bit map
How to Build Your Own Kibana Plug-ins
(Tanya Bragin and Spencer Alger)
This was a very interesting presentation with some live hacking of creating a custom plugin showing how to create a new visual component in Kibana. The plugin takes a new D3 Guage component and makes it available in Kibana.
There was a lot of code in the presentation that is going to be put on github. We have a lot of interesting ideas for Kibana plugins, so stay tuned for interesting new blog post in the near future.
Bringing Healthcare Analytics to the Point-of-Care @ Mayo Clinic
Interesting story about using elasticsearch to store medical events in elasticsearch and do analysis over a lot of patients to predict what could happen to current patients. They used technologies like the percolator as well as parent child relation ships to enrich the events before actually being stored.
Quantitative Cluster Sizing
(Christian Dahlqvist and Ryan Schneider)
This session did not give a lot of insights in how to size your cluster, but it did explain why. Also it showed how you can do experiments with your own data and how to evaluate the results to make it possible to determine the number of shards, size of hardware for your situation.
The conference ended with a chat between Shay CTO from elastic and Lew CTO cloud from Cisco. They talked about open source and the future of it. Finally Steven took the stage to thank for attending the conference and to ask if we want to come back next year.