Skip to main content
Version: Python

Why do some queries run faster the second time they are run?

Deephaven utilizes memoization for many table operations, such as avg_by, which stores and recalls previously computed results. This optimization significantly reduces execution time on repeated runs.

Are all operations in Deephaven memoized?​

Not all operations are memoized. Some operations are excluded for one of two reasons:

  • The work hasn't been done to memoize the table operation (yet).
  • Memoization is unsafe, especially with stateful User-Defined Functions (UDFs).

Are there other factors contributing to improved execution time on subsequent runs?​

Yes. Other factors include the following, which is often done on the first pass:

  • Warmed filesystem and/or operating system buffer caches, which result in a complete in-memory cache of materialized pages.
  • JIT compilation, which optimizes execution for subsequent runs.

Can memoization be disabled for benchmark testing?​

Yes, if you're benchmarking and want to disable memoization, you can do so. Set the property QueryTable.memoizeResults to false.

You can include this setting in your .prop file:

QueryTable.memoizeResults=false

Or as a JVM argument:

docker run --rm --env START_OPTS="-DQueryTable.memoizeResults=false" ghcr.io/deephaven/server:latest
note

These FAQ pages contain answers to questions about Deephaven Community Core that our users have asked in our Community Slack. If you have a question that is not in our documentation, join our Community and we'll be happy to help!