Thought of the day: After more than 7 years of focusing, as developer, mainly on designing code for large games as a service, I have not much doubts left about how to write proper architectures. With this wisdom I can say that DOD is, at the moment, the best paradigm to follow...
-
-
DOD may deliver 10x per thread or 100x per CPU at enormous cost in coding overhead and fragility, but 1000x to 1,000,000x have to come from distributed computing of some sort.
-
The transaction experiments are aimed at exploring the later magnitudes of gains by scaling with brute force runtime techniques that don’t diminish the ease of coding. However, feasibility is still far from established.
End of conversation
New conversation -
-
-
These said, I'm not sure if single box performance has to be sacrificed if we want the main loop to be async and distributed, in fact ECS might ease the transition to that if you write the systems with async component modifications which can fail but yes that is hard to code for
-
It's been a big learning curve for us but we'll be releasing a sim game next year that is written from the ground up with multithreaded-by-default ECS. Time will tell if it was all worth it!
- 1 more reply
New conversation -
-
-
Yes back then when at MuchDifferent we were still a company we tried to do it by putting a distributed on top of single threaded Unity and UE4 processes a bit similar to improbable's approach. Was called PikkoServer and we were close to breaking records
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Bandwidth saturation between internal nodes and cost of coordination were too high and I think writing the game in a distributed async manner from the get-go might be the only way to achieve that but gameplay programming then is anything but straightforward
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I'm fully aware that still some write/modifications will need 2PC and some systems might need coordination of some forms or we need actor model like requests between instances of a system on different machines and that might mean not much perf gains
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.