Also worth looking at this latest video of Worlds Adrift, one of the games built on SpatialOS, that is now in Alpha: https://www.youtube.com/watch?v=kjLWK6iQvRs Beautiful storytelling on this one :)
Though different terms are used, both SpatialOS and "Age of Ascent" specifically reference the notion of "Interest Management." That is, not only do the online servers have to simulate the world at the rate of game ticks, the online servers also have to calculate and generate timely updates to interested parties and compensate for lag while seamlessly distributing load across the cluster.
3D worlds are really good for interest management, since distance between objects generally means a drastic lowering of bandwidth required between the objects, even though the objects are still visible to each other. In 2D, such objects have zero bandwidth, but this means there are no "vistas" unless the view zooms out, which makes the player effectively "farther" from the avatar.
why would I entrust the future of my company or game project to a closed source server and SDK?
I've seen several MMO middleware companies announce revolutionary new technology with outdated closed source business models only to fold couple of years layer. Badumna Network suite was the last well funded one I remember closed a while back. And it started by Australian academics with decent technology.
I had a terrible time parsing the title (currently "Improbable Partners with Google, and Announces SpatialOS Game Developer Alpha") until I realised that 'Improbable' was the name of the company, and 'Partners' a verb.
(disclaimer: I'm a Cofounder & CTO of Improbable - and massive fan of Second Life)
You're right to see that there are some inspirations from worlds like Second Life with our platform - it spatially distributed the simulation of the world through multiple servers.
Will elaborate on this in a blog post a some point, but some main differences:
-SL is a great example of a product somebody could make on top of SpatialOS: we're less of a 'social world environment', and more of an abstract spatial computation platform. We don't care what game engine is used, what the view of the world is, etc.
-While really impressive for its time, SL's architecture didn't re-distribute its load across machines based upon player & object density - each server was statically allocated a part of the world. This meant concurrent players per 'Sim' were limited to ~50. We can move these boundaries on the fly to deal with higher density. Here is an example (we really need a better one, this doesn't do it justice!) https://www.youtube.com/watch?v=XXhxI4dKU_E&t=1s
-Although a player could cross server boundaries, it was never very smooth, and people in SL would deliberately limit their playspaces to a single server. To create a proper seamless world you have to deal with multiple servers simulating the same physical space, and reconciling with each other. Here is an example of us doing this with Unity3D and physics https://www.youtube.com/watch?v=ysB6bi7QUbw
I'm curious how SpatialOS compares to Second Life's successors -- Sansar and High Fidelity. I gather that they also allocate processing in a more targeted way, to respond to demand.
P.S. I'm also a huge fan of SL, and look forward to trying out whatever similar world emerges using SpatialOS.
I'm not sure what they were doing a decade ago, but this seems to be a more ubiquitous task management system for multiplayer with any game across multiple platforms.
Linden Labs left things a bit too free-form. What happens when you let people do what they want in a virtual reality? They go up the Maslow Need Hierarchy.
Food, safety, shelter, etc aren't an issue, so they zoom up the needs hierarchy and stop at sex. There needs to be an overarching theme to focus people. Online worlds need to have needs. Otherwise, they can't get too far beyond being a glorified Garry's Mod server. It's like a musical jam session with no structure at all. That kind of thing only works if people already know each other. In public, it's always too chaotic.
I've seen this kind of middleware come through several times. It's a hard sell because MMOs are already big productions and this stuff, while technically very challenging, only captures a single bullet point feature: "world is seamless and persistent." And from the player POV, that is cool, but not I'll Buy That cool, especially when scale ends up creating all sorts of other design problems(more players and bigger worlds = more potential for chaos, harder UX).
Our Platform team uses Golang extensively. We use it together with gRPC to form our microservice stack that orchestrates the SpatialOS simulations.
The TL of our Platform API team did a cool talk at the Go London User Group meetup in October: https://skillsmatter.com/skillscasts/9259-a-debuggable-base-...
Very interesting. I have a multiplayer game I'm working on.
http://www.emergencevector.com
The servers are written in Golang. There is a "Master Control Process" that acts as a coordinator, and a farm of worker processes that can report on their population/load and host star-system instances. (Basically, a game loop on an R-Tree.) The coordinator just uses the best available information and assigns clients to hosts. Clients join hosts in an idempotent fashion, so it doesn't make any difference if an instance currently exists or not. Each process is structured somewhat like a Erlang actor.
I might try to do a port of my game on your servers, just out of curiosity.
That's interesting, we have a slightly related architecture. SpatialOS Runtime, itself distributed across many machines, acts as your "Master Control Process". It orchestrates Workers, both managed (running in the cloud), and external (remotely connected game clients), to perform the co-simulation of the world.
The key feature of the stack is that you can integrate various Workers to act together, leveraging existing game engines in multiple languages. Checkout https://spatialos.improbable.io/docs/reference/9.0-alpha/wor... if you're interested in trying to port your stuff onto our stack.
How long before you let people write workers in Golang?
My equivalent to SpatialOS is not only my MCP (Tron reference, of course) but also a bit in each worker process. What you call Workers are what I call Instances. We both have pretty much the same notion of Entity. However, I don't have External Workers. Instead, I have a "Pilot" object for every "Ship" (Ship == Entity) and one kind of Pilot represents the game Client. The difference makes sense, as you have a notion of Workers sharing data, so your Clients are external to an Entity.
We have a C Worker API (and in fact the C++, C# and Java worker APIs are built on top of it), so you can write workers in any language that can call C functions. It is a non-trivial amount of work, though -- see https://spatialos.improbable.io/docs/reference/9.0-alpha/wor... to see what one of these APIs look like.
For the moment we don't have the SDK supporting Go but we are definitely taking input on such topics. You should definitely post this request on the forum as there will probably be other people wanting the same thing! Will be happy to discuss it further there.
The idea of a Go SDK has already crossed some minds at Improbable for certain.
Hi, thanks for doing the tutorial! It looks like you may need to run `spatial init` and then try `spatial update` again?
If this doesn't work, please ask your question on https://support.improbable.io/ -- there are many more people there to provide support and share ideas.
Also worth looking at this latest video of Worlds Adrift, one of the games built on SpatialOS, that is now in Alpha: https://www.youtube.com/watch?v=kjLWK6iQvRs Beautiful storytelling on this one :)
The website (https://spatialos.improbable.io) has more details about the tech itself, and for an even more in-depth look, the documentation (https://spatialos.improbable.io/docs/reference) has everything from a conceptual overview to tutorials to API references.