London-based Improbable unveils SpatialOS for distributed simulation | Hacker News Reader

London-based Improbable unveils SpatialOS for distributed simulation

alphr.com

68 pointsggambetta25 comments

I am a bit sceptical about this startup.

Two friends interviewed there and both weren't impressed.

One of them is the best developer I know, with decades of cutting edge AAA games engine development experience, including in the MMO space and his words were: "they have no clue about games development and don't have much more than vapourware"

I hope he is wrong though and that the company is successful. They seem to be one of the few games related companies to compete with finance on compensation packages in London.

Interesting, did he mention any specifics? They seem to be aiming to create a general platform that's not specific to games. They also do have a few marketing / business managers which might be why it sounded like vapourware? As far as I can tell they have a lot of talent in the engineering department.

Sounds more like Virtual Software/Infrastructure as a Service than a conventional OS:

Unsurprisingly, working in SpatialOS will require special skills. Games can still be built with the Unity engine, though they'll also need to be able to work with Scala, the native SpatialOS language. However, Narula said the World's Adrift team needed "very little" training. From there, it's simply "pushed" out onto Improbable's infrastructure, and congratulations, you're a maker of worlds.

Cost Structure:

Narula wouldn't reveal pricing, but you only pay for usage – if no one is in your world, there's no cost – and he suggested Bossa's ability to build Worlds Adrift shows it's affordable.

If I'm understanding this correctly, it's more that they're building large, persistent "environments" that can easily integrate with existing development and modeling software. Virtual sandboxes for playing and testing.

Anyone have enough info to provide clarification?

> need to be able to work with Scala

This kills the startup.

Disagree. Being able to work with Scala doesn't mean you have to write Scala always (depending on how they structured/exposed it). And even if it does mean that now, since it's on the JVM, it is not an unreasonable effort to create language-specific facades. Saying something "kills the startup" is extreme.

Are you suggesting not supporting a language thats not in top 25 languages today is their #1 problem?

I think he's suggesting that only supporting a language thats not in top 25 languages today is their #1 problem

Nothing new under the sun - this sounds just like Ab Initio Co>OS (https://www.abinitio.com started somewhere in 1995 and I would say the best piece of software code I've seen in my 20 years - not working for them anymore, so it is not advertisement).

Well, I had a look at their website.

It takes twenty seconds to load. (I timed it.) And all that delay is deliberate; it's stupid animations and scrolly things, all of them unskippable. Once it's loaded you now have a different stupid scrolly thing which is almost impossible to navig...

Oh god. It's flash. The whole site is one huge flash page! What is this, the 1990s?

If their marketing and PR department is so utterly out-of-touch that they think this is a good idea, I shudder to think what the engineering's like!

Gamasutra has a more gaming focussed article on it: http://www.gamasutra.com/view/news/256015/Improbable_How_a_t...

MMORPG backend as a service?

Better summary than TFA

So... it's Second Life as a service.

best TL;DR

The hype all sounds well and good, but it seems to be what they're really offering is an opportunity to do "large scale" monte-carlo simulations. Which is also a great thing, but the results of said simulations will only be as good as the detail to which they are modeled.

The example that they give about utilities modeling usage in a city is great, but someone is going to spend a whole lot of time modeling the details of the city, and constantly verifying that those results match up with the real world.

I spent a lot of time doing these kinds of simulations in school...

From an chat with one of their employees, my understanding is that they have a nice SDK to make it easy for people to write simulations thay scale well, by enforcing some constraints:

Dependencies between agents are local and can be expressed by simple messages

Low dimensional and smooth action space

Simple per-agent dynamics

From what I've heard it isn't going to revolutionise anything, but it might make more massive open world space MMORPGs which I'm strongly in favour of.

First time I saw a post about these guys I said that they're just using the actor pattern...now I hear that they use Scala...another point for actors!

I'm surprised that none of the companies in this space are applying this sort of technology to simulating financial markets.

The point of this project/company is that the agents are spatially embedded. Markets are almost abstracted from their physical layout (with the exception of high-frequency stuff where the speed of light breaks the abstraction).

There is a big difference in computational requirements between simulating a system and constructing/analyzing a model of a system from measurements, doubly so if the system being modeled is live. To put it another way, you typically can't use virtual world platform computer science to build a mirror world platform that captures complex dynamics happening in reality. Very different problems.

For financial markets, modeling and analyzing systems as they are measured is far more useful (and what is done in practice), so spatial simulation platforms are not that interesting.

> Simulated worlds provide unique insight to those asking questions of complex systems. As well as enabling completely new experiences in gaming, simulated worlds can solve significant problems in areas as diverse as defence, energy, city efficiency, health, and finance.

https://angel.co/improbable

They'll get to it trust me...

I signed up for early access, and they asked what business I was in. Choices included games, academia, and finance.

Is there technical information on this somewhere? How is this different from AWS and from a traditional supercomputer? If I write a shared-memory parallel computation could I just run it on their platform (assuming I've done the work to make sure it can scale), or do I need to use MPI/a special language to make it all work?

There's more information on the website: http://improbable.io/learn-more

Looks like you need to use Scala for this.

As for the comparison with supercomputing: this falls decidedly in the abstracted/virtualized/cloud category, which is basically the opposite of a supercomputer, where you're very close to the metal (i.e. C/C++ or Fortran). Efficient supercomputer usage requires hard thinking about communications (both between nodes with MPI and between the RAM and the CPU on each node), avoiding cache thrashing at all cost, and making use of all the performance intrinsics you can.

Reply on news.ycombinator.com