10.2. Tuning Presto

The default Presto settings should work well for most workloads. The following information may help you if your cluster is facing a specific performance problem.

When setting up a cluster for a specific workload it may be necessary to adjust the following properties to ensure optimal performance:

distributed-joins-enabled

query.max-memory

query.max-memory-per-node

query.initial-hash-partitions

task.concurrency

Those and other Presto properties are described in properties article.

As an example, on a 11-node cluster dedicated to Presto (1 Coordinator + 10 Workers) with 8-core CPU and 128GB of RAM per node, you might want to start tuning with following values:

query.max-memory = 200GB

query.max-memory-per-node = 32GB

query.initial-hash-partitions = 10

task.concurrency = 8

If your workload consists of queries that have very quick processing time of the splits you might encounter situation when whole cluster is under utilized (very small usage of CPU, network and disks). For example this can happen when your queries are highly selective and most splits can be filtered out just by looking into theirs headers (like bloom filters or min/max statistics in ORC files) without reading the actual data. In such case you might want to increase values of the following properties:

node-scheduler.max-pending-splits-per-node-per-stage

node-scheduler.max-splits-per-node

If this guide does not suit your needs, You may look for more tuning options on Properties Reference page.

JVM Settings

The following can be helpful for diagnosing GC issues:

-XX:+PrintGCApplicationConcurrentTime
-XX:+PrintGCApplicationStoppedTime
-XX:+PrintGCCause
-XX:+PrintGCDateStamps
-XX:+PrintGCTimeStamps
-XX:+PrintGCDetails
-XX:+PrintReferenceGC
-XX:+PrintClassHistogramAfterFullGC
-XX:+PrintClassHistogramBeforeFullGC
-XX:PrintFLSStatistics=2
-XX:+PrintAdaptiveSizePolicy
-XX:+PrintSafepointStatistics
-XX:PrintSafepointStatisticsCount=1