I was thinking that instead of having say a game which rendered entirely in the cloud or entirely locally that instead the game could render in a mixed mode where some of the computationally heavy tasks are offloaded to the cloud whilst the latency sensitive tasks were kept local. I think this could be a much smarter model than to render either entirely locally or entirely in the cloud.
Say you've got a game like Halo 5 running on Durango. If you're playing offline or your internet goes down you play the game in local mode which runs entirely off your Durango box. However if you've got a live internet connection instead the game offloads some of the computation work onto the Xbox Live servers. The question I have is this a viable model in terms of both development work required to implement as well as practical considerations given the number of computers? It seems more practical to do some of the rendering on the cloud than doing all of it and it could be easier to load balance because if you could scale the workload so that if the load gets heavy the cloud component would do less work for each client.
So what parts of the workload could be done on the cloud?
I was thinking:
1. Lighting. I figure that lighting could be done offline as well as online because it seems computationally intensive but not data intensive. It could be scaled between say having a fixed lightmap baked in or completely dynamic lights.
2. AI. I figure this is another interesting one as again this is computationally intensive and yet it can scale between being done locally with say dumb AI and being done on the cloud with much smarter AI. This is also latency insensitive as having slightly slower reacting AI wouldn't be a huge negative.
3. Voice processing. Apple has already proved that this is a viable model with Siri, Microsoft could do something similar as well.
4. Server hosting. I think this one is pretty obvious, hosting multiplayer sessions is pretty intensive computationally.
5. Environmental physics. Things like destructibility and how things collapse or deform could be done offline. Also things like soft body physics like for instance flag waving as well as hard body physics, anything which isn't directly related to player actions could be done in the cloud.
6+. Anything you guys can come up with.
I think it would be a great way to use those Live payments people make as well as a great way to offer high performance whilst turning something which was once a source of losses into a source of profit by letting people subscribe to extra performance. I think as a business case it is justifiable which leaves only technical reasons as to why it would/would not work.
Say you've got a game like Halo 5 running on Durango. If you're playing offline or your internet goes down you play the game in local mode which runs entirely off your Durango box. However if you've got a live internet connection instead the game offloads some of the computation work onto the Xbox Live servers. The question I have is this a viable model in terms of both development work required to implement as well as practical considerations given the number of computers? It seems more practical to do some of the rendering on the cloud than doing all of it and it could be easier to load balance because if you could scale the workload so that if the load gets heavy the cloud component would do less work for each client.
So what parts of the workload could be done on the cloud?
I was thinking:
1. Lighting. I figure that lighting could be done offline as well as online because it seems computationally intensive but not data intensive. It could be scaled between say having a fixed lightmap baked in or completely dynamic lights.
2. AI. I figure this is another interesting one as again this is computationally intensive and yet it can scale between being done locally with say dumb AI and being done on the cloud with much smarter AI. This is also latency insensitive as having slightly slower reacting AI wouldn't be a huge negative.
3. Voice processing. Apple has already proved that this is a viable model with Siri, Microsoft could do something similar as well.
4. Server hosting. I think this one is pretty obvious, hosting multiplayer sessions is pretty intensive computationally.
5. Environmental physics. Things like destructibility and how things collapse or deform could be done offline. Also things like soft body physics like for instance flag waving as well as hard body physics, anything which isn't directly related to player actions could be done in the cloud.
6+. Anything you guys can come up with.
I think it would be a great way to use those Live payments people make as well as a great way to offer high performance whilst turning something which was once a source of losses into a source of profit by letting people subscribe to extra performance. I think as a business case it is justifiable which leaves only technical reasons as to why it would/would not work.
