DZone
Thanks for visiting DZone today,
Edit Profile
  • Manage Email Subscriptions
  • How to Post to DZone
  • Article Submission Guidelines
Sign Out View Profile
  • Post an Article
  • Manage My Drafts
Newsletter
Log In / Join
Refcards Trend Reports
Events Video Library
Refcards
Trend Reports

Events

View Events Video Library

Related

  • Does a Bigger Context Window Mean Better Memory: Understanding Context Windows, Model Memory, and RAG
  • How Cache Invalidation Caused Memory Pressure and API Failures at Scale
  • Remember Me, Safely: Durable and Governed Memory for Enterprise AI Agents
  • Why Incident Response Needs Memory, Not Just Intelligence

Trending

  • YAGNI in the Age of AI Coding
  • How to Migrate From LangChainGo to Genkit Go
  • Should Developers Switch From SQLite to DuckDB?
  • Searching Kafka Without an Index: Lessons From Building Message Search for Kafdrop

Draft This article hasn't been published, so readers can't see it.

In moderation This article is waiting for review, so readers can't see it yet.

Deleted This article has been deleted, so readers can't see it.

MineCraft and Off Heap Memory

By 
Peter Lawrey
·
Jun. 19, 14 · Interview
Post on X (opens in a new tab) Share on LinkedIn (opens in a new tab)
7.6K Views

Join the DZone community Save articles, join the discussion, and download Refcards and Trend Reports with a free account.

Join for free Already a member? Log in
MineCraft is a really good example of when off heap memory can really help.
The key requirement are
  • The bulk of the retained data is a simple data structure (in minecraft's case its lots of byte[])
  • The usage of off heap memory can be hidden in abstraction.

The test

I used the following test for starting minecraft server from a seed from scratch which is a particularly expensive operation for the server.
  • Preset the level-seed=114 in server.properties
  • Delete the world* directories
  • Start the server with these options to see what the GC is doing -Xloggc:gc.log 
    -verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps 
    -XX:+PrintGCTimeStamps -XX:+PrintTenuringDistribution -Xmx6g
  • Connect with one client
  • Perform /worldgen village
  • Perform /save-all
  • Exit.
To analyse the logs I am using jClarity's Censum.

Standard MineCraft

There is two particularly expensive things it does
  • It caches block stage in many byte[]s
  • It attempts to cache int[] used for processing without limit.
A Censum report for the above test looks like this
The high pause times are partly due to having to manage the large objects.

Off heap MineCraft

Two changes were made to address this
  • Use off heap ByteBuffer for long term caching.  Unsafe would be more efficient but not as portable.
  • Put a cap on the number of int[] cached.
Note: the problem with the temporary int[] only became visible to me after moving the bulk of the data off heap. Addressing the biggest problem over reveals more quick fix problems.
A Censum report for the same test looks like this
There is still some premature promotion i.e. further improvements can be made but you can see that the application is spending.

Conclusion

Using off heap memory can help you tame you GC pause times, especially if the bulk of your data is in simple data structures which can be easily abstracted.  Doing so can also help reveal other simple optimisations you can do to improve the consistency of your performance.

Footnote

Many organisation treat performance as an enhancement and optional, however if you instil a culture where reasonable performance is a requirement, and failing to meet this requirement is a bug, performance issue are more likely to be fix. c.f.https://bugs.mojang.com/browse/MC-56447

The Source Used

The source used for the test is available here. https://github.com/peter-lawrey/MineOffHeap
The logs produced are available here. https://github.com/peter-lawrey/MineOffHeap/tree/master/logs
Memory (storage engine)

Published at DZone with permission of Peter Lawrey. See the original article here.

Opinions expressed by DZone contributors are their own.

Related

  • Does a Bigger Context Window Mean Better Memory: Understanding Context Windows, Model Memory, and RAG
  • How Cache Invalidation Caused Memory Pressure and API Failures at Scale
  • Remember Me, Safely: Durable and Governed Memory for Enterprise AI Agents
  • Why Incident Response Needs Memory, Not Just Intelligence

Partner Resources

×

Comments

Loading comments

The comments didn't load

Please try again in a moment.

Likes

Loading

The next likes didn't load. Please try again.

The likes didn't load

Please try again in a moment.

See who liked this

Log in or join DZone for free to see who liked this article.

Log in Join for free
  • RSS
  • X
  • Facebook

ABOUT US

  • About DZone
  • Support and feedback
  • Community research

ADVERTISE

  • Advertise with DZone

CONTRIBUTE ON DZONE

  • Article Submission Guidelines
  • Become a Contributor
  • Core Program
  • Visit the Writers' Zone

LEGAL

  • Terms of Service
  • Privacy Policy

CONTACT US

  • 3343 Perimeter Hill Drive
  • Suite 215
  • Nashville, TN 37211
  • [email protected]

Let's be friends:

  • RSS
  • X
  • Facebook