An Architecture for Millions of Things

Re:An Architecture for Millions of Things by yetihehe on Thursday July 14, 2016 @02:33AM (#12092028)

If you need to do something which was never done because of efficiency problems, you need to think about how to do it before you start. It's called planning, not premature optimization. Just optimizing existing approaches won't do anything good here as author said in beginning.

[ Parent | on HN ]

Re:An Architecture for Millions of Things by ggambetta on Thursday July 14, 2016 @07:00AM (#12092811)

It is a bit premature in a way; paraphrasing here, but "a million is not cool. You know what's cool? A billion" No matter how much you optimise CPU caches and memory allocators, at some point your application won't fit in a single machine, and you'll have to go distributed. Datacenters, not mainframes :)

[ Parent | on HN ]

Re:An Architecture for Millions of Things by anselm_eickhoff on Thursday July 14, 2016 @08:22AM (#12093122)

But if you're already using message passing, this transition will be much more natural!

[ Parent | on HN ]

Re:An Architecture for Millions of Things by Dzugaru on Thursday July 14, 2016 @06:32AM (#12092721)

> which was never done

Highly doubt that.

Oh, and when I see a phrase "millions of things interacting" and don't see a word "GPU" anywhere I feel strange. Just saying.

[ Parent | on HN ]

Re:An Architecture for Millions of Things by majewsky on Thursday July 14, 2016 @07:43AM (#12092967)

Assuming that he is talking about a city sim as a game, I can understand that he doesn't want to use the GPU for the simulation, since it will be occupied with the (even more expensive) rendering of the city.

[ Parent | on HN ]

Re:An Architecture for Millions of Things by anselm_eickhoff on Thursday July 14, 2016 @08:12AM (#12093078)

That and that I really want to avoid having to debug GPGPU code for the main simulation. But I'm not saying it won't become GPGPU at some point!

[ Parent | on HN ]