Elasticsearch Tuning 202

Upcoming Elasticsearch Classes 2019

Mar 12, 2019$200 / person Only $180 / person before Feb 6thRegister Now
May 14, 2019$200 / person Only $180 / person before 6th AprRegister Now
Sept 24, 2019$200 / person Only $180 / person before 20 JulyRegister Now
Dec 10, 2019$200 / person Only $180 / person before 30 SeptRegister Now


Radu Gheorghe

This class teaches you how to tune Elasticsearch low-level settings, from merges and thread pools down to hardware choices, in order to have fast, and therefore easier to scale clusters. Radu Gheorghe, a seasoned Elasticsearch instructor, and consultant from Sematext, author of “Elasticsearch in Action”, and frequent conference speaker will be your trainer.

Who Should Attend

This course is designed for anyone with a good understanding of Elasticsearch core concepts (indices, shards, replicas, nodes), who wants to understand the low-level performance tuning knobs of Elasticsearch. From the JVM-level GC options, to Lucene-level index sorting and much more.

Why Attend

Using hardware efficiently impacts both the cost/performance ratio of your deployment and how much you can scale it without too much administrative overhead. This class will get you deep into how Elasticsearch and Lucene organize data on disk while reading and writing, uses threads and what are the typical hardware bottlenecks, so you can tune your cluster for better performance. Further benefits:

  • A customized learning experience, targeted for solving specific use-cases
  • Classes are instructor-led and exercises are derived from years of working with clients.
  • Small class sizes allowing for more interaction and more time to discuss what matters to you in practice.
  • More flexible – no need to travel, a short class is easier to fit in your schedule
  • Same high-quality instruction as our public or private Elasticsearch classes

Course Outline

  • Choosing the right hardware for the job: index, search, aggregations, masters or clients
  • Tuning the merge policy to balance index and search performance, while avoiding spiky load
  • Tuning GC settings to make big clusters faster and more stable
  • Using thread pools to reserve CPU and memory for important workloads
  • When does index pre-sorting help and when it just adds overhead
  • Lab
    • Tune the thread pool to allow more space for searches
    • Using the G1 GC to get more throughput and lower latency on larger heaps
    • Tweaking the merge policy for heavy writes
    • Activating index sort

Things to Remember

For the online training, all participants must use their own computer with OSX, Linux, or Windows, with the latest version of Docker installed. A modern browser is needed to join the virtual classroom, as well as a working headset, microphone and camera.

Participants should also be comfortable using a terminal/command line. Sematext provides:

  • A digital copy of the training material, including exercises
  • An archive with Docker Compose files, configs, scripts, etc.