Meet Improbable, the Startup Building the World's Most Powerful Simulations(forbes.com)
forbes.com
Meet Improbable, the Startup Building the World's Most Powerful Simulations
http://www.forbes.com/sites/parmyolson/2015/05/27/improbable-startup-simulations/
12 comments
Yes, Second Life has one persistent world ("one shard") but it's composed of a bunch of "sims" glued together at sim boundaries. (A "sim" is a 256 m x 256 m x notsure m volume of space.) The states of the things in that sim are the responsibility of one server; if an object crosses a sim boundary, the new sim may be handled by a different server. The sim-crossing hand-off can cause object motions to look non-physical.
As I understand it, Improbable also simulates one big world, but they map the various simulation calculations to servers in a different way. How exactly? That's their secret sauce.
Philip Rosedale's new startup, High Fidelity, also takes a different approach. See https://highfidelity.com/
As I understand it, Improbable also simulates one big world, but they map the various simulation calculations to servers in a different way. How exactly? That's their secret sauce.
Philip Rosedale's new startup, High Fidelity, also takes a different approach. See https://highfidelity.com/
A bit more information can be found here: http://cdixon.org/2015/03/24/improbable-enabling-the-develop...
Too me it reads like MapReduce for highly place-dependent computations, whatever that looks like. Probably something along the lines of a distributed kd-tree with message passing at borders handled for you as well.
Too me it reads like MapReduce for highly place-dependent computations, whatever that looks like. Probably something along the lines of a distributed kd-tree with message passing at borders handled for you as well.
Maybe that's why they showed a flying pirate scenario.
Imagine every object sitting on the same planet (object). Suddenly, everything interacts with everything. You can't know if the object is a bee gently landing on a flower, or a 20 mile radius asteroid crashing into earth. The effect of the latter would be felt everywhere.
So, if every object was sitting on the same planet (object), even their system would probably crash and burn with n objects passing n^2 messages all the time.
The kd-tree idea sounds neat. They would need to adapt it however, to allow objects to move (to pass in and out of other objects' interaction radii).
Imagine every object sitting on the same planet (object). Suddenly, everything interacts with everything. You can't know if the object is a bee gently landing on a flower, or a 20 mile radius asteroid crashing into earth. The effect of the latter would be felt everywhere.
So, if every object was sitting on the same planet (object), even their system would probably crash and burn with n objects passing n^2 messages all the time.
The kd-tree idea sounds neat. They would need to adapt it however, to allow objects to move (to pass in and out of other objects' interaction radii).
Improbable's tech intrigues me, but I agree with many of the comments that technical information on how the actual tech works is limited, and most of the articles are puff pieces that focus on aspects of Bossa Studio's game that have nothing to do with Improbable, the part about emergent gameplay for example. According to Glenn Fiedler (www.gafferongames.com) in his recent GDC talk on networking physics (http://gdcvault.com/play/1022195/Physics-for-Game-Programmer...) this paper by Insomniac Games (https://d3cw3dd2w32x2b.cloudfront.net/wp-content/uploads/201...) details a system similar to what Improbable is working on.
In this system the server is basically a distributed database and message passing system, and grants different clients authority over specific simulation objects or fields. If a client is the authority it runs the simulation logic for that object. As long as you can scale you message passing and db it seems like most work can be offloaded on clients, so complex simulation is "free". I'd be concerned about cheating in more competitive games, but it should work well for social or creative games and for non-game simulation where you know clients are trustworthy.
In this system the server is basically a distributed database and message passing system, and grants different clients authority over specific simulation objects or fields. If a client is the authority it runs the simulation logic for that object. As long as you can scale you message passing and db it seems like most work can be offloaded on clients, so complex simulation is "free". I'd be concerned about cheating in more competitive games, but it should work well for social or creative games and for non-game simulation where you know clients are trustworthy.
Here's the one-page article - http://www.forbes.com/sites/parmyolson/2015/05/27/improbable...
Splitting this article into 5 pages to increase page views is really sad & completely unneeded.
Splitting this article into 5 pages to increase page views is really sad & completely unneeded.
Does it not at least have the added benefit of giving them the metric of "kept reading article till page 3, then left". Even if they just infer that there was something wrong with page 3, rather than the user getting peeved-off that they have to keep paging to read the rest of the story.
periodically sending the scroll position to the server is a much better alternative
vapid article but more than that the UI for actually getting through the story is awful. the recommended articles box above the much smaller 'Continue' button, is particularly egregious.
Yeah I'm disappointed in Forbes. Their new website interface is garbage, practically unusable.
Here is the website of the game itself: http://www.bossastudios.com/games/worlds-adrift/
Here is the website of the game itself: http://www.bossastudios.com/games/worlds-adrift/
thanks for that.
I just googled 'Improbable London'. On their website not much substantive , just a standard 'hello world' post: http://improbable.io/blog/
I'm personally interested in such an approach for an academic tasks ( complex electrical simulations), but I have no idea where to find some substantive discussion of what they're doing here.
kind of sad
I just googled 'Improbable London'. On their website not much substantive , just a standard 'hello world' post: http://improbable.io/blog/
I'm personally interested in such an approach for an academic tasks ( complex electrical simulations), but I have no idea where to find some substantive discussion of what they're doing here.
kind of sad
Ironic that this talks about modeling games for business and other use cases when I read http://www.perworks.com/my-son-has-23-6-billion-how-is-yours... just a few minutes before on Hackernews. (the link is about building business models as games letting them loose and using the wisdom of the crowds to make decisions).
It's worth noting that large-scale simulations are nothing new in the world of science. For example, consider the EAGLE simulation of the whole universe with realistic galaxies: https://www.ljmu.ac.uk/about-us/news/ari-creates-simulation-...
Improbable seems to be different in that it's aiming at the gaming market? But apparently they also have science clients? Maybe the difference is in their computational architecture, or cost?
Improbable seems to be different in that it's aiming at the gaming market? But apparently they also have science clients? Maybe the difference is in their computational architecture, or cost?
"The game was also built for a fraction of the usual cost. ... would typically take years and millions of dollars for a studio of Bossa’s size to build, but they did it in roughly a year and with just a core team of front-end developers."
vs
"In their ground floor office on a bland-looking block in Farringdon, a team of about 60 engineers from MIT, Goldman Sachs and Google sit at $40 desks writing code..."
Typical in gamedev :) (e.g. Unity Technologies):
"“Eventually we realized the tech we were working on was bigger than the game,” says Whitehead."
vs
"In their ground floor office on a bland-looking block in Farringdon, a team of about 60 engineers from MIT, Goldman Sachs and Google sit at $40 desks writing code..."
Typical in gamedev :) (e.g. Unity Technologies):
"“Eventually we realized the tech we were working on was bigger than the game,” says Whitehead."
They look incredibly nice and skilled though. Too bad I don't possess a useful skillset nor an outstanding pedigree.
Anyway, reporting verbatim their CEO "We have created a world where there are no server boundaries", so I assume they managed to scale the instancing method seen in Guild Wars 2 (where overflowing players are divided among different ghost instances) into a single giant concurrent world.
Just becasue I learned about Worlds Adrift I'm glad I read this article.
They often focus on the "kill a monster and it stays dead" angle, so that seems like a real aspect of the system, but it's not an advance. Other companies have aimed at this "persistent virtual world-building toolkit" target before, from Second Life to Blue Mars to Metaverse to Metaplace. What distinguishes Improbable from these other ideas?