Instance streaming best practices and techniques

Hi Creators!

We just published a new guide covering some of the best practices and techniques for instance streaming, so you can more effectively use it in your experiences.

:star: Check out the Instance Streaming Techniques page!

Below is a quick look at some of the existing streaming resources we’ve released along the way.

A Look Back

:rocket: SLIM (Scalable Lightweight Interactive Models) is a system that works hand-in-hand with instance streaming to automatically generate optimized, lightweight representations of your models when they’re streamed out. Instead of models simply disappearing at a distance, SLIM composites them into efficient low-poly versions that preserve visual quality across all device types.
Read more about SLIM in these posts: SLIM Beta and Introducing SLIM.

:television: This video explains ways to manage the engine’s rendering detail based on player distance from 3D objects in your experience.

:television: Also check out how instance streaming and other factors can affect the engine’s memory usage and performance:

:cop: Server Authority is a networking model that is built on instance streaming to make the server the single source of truth for your experience’s state. This is particularly valuable for competitive games like first-person shooters, racers, and sports experiences, where client-side exploits like speed hacking or teleportation have historically been difficult to prevent.
Read more about server authority in this post: Server Authority Beta

:books: As always, the core instance streaming doc is a great companion if you want to learn more about streaming. Check out the Instance Streaming Overview page.

Let us know what you think

We’d love to hear your feedback on this new guide. Let us know if we missed anything or if there are topics you’d like us to cover in future content. Drop a comment below and happy creating!

13 Likes

I thought there was a new video. :confused:

 

Streaming Enabled wisdom though

If you have a Model with a bunch of Parts. Maybe it’s better to keep it on Persistent or something, because everytime you load that Model back… there’s stutter.

For example, take a boat and go through all the islands in FISCH. Which while you do that you pass by some islands, which try to load as well, and yep.

The result:

vs. while I stay in place and do nothing

Less green bars. I think it has to be a Streaming Enabled thing.

 

What is SmoothClusterTaskQueue ?

But in the MicroProfiler what I’d see is Replicator ProcessPackets.

3 Likes

Howdy!
Thanks for the input, and you are correct: large atomic models can be performance bottlenecks as they must replicate all at once. This is actually mentioned in the best practices doc above in Organize Models Appropriately. Models with high part counts are generally discouraged, but we understand they may be unavoidable. We are also aware of this performance issue and have plans to alleviate such.

4 Likes