Server based game augmentations. The transition to cloud. Really possible?

Personally, I think it's more useful to split the cloud into it's 2 aspects:
- compute cloud. High latency, high performance, long-running processes. Suitable for questions like 'what is the meaning of life', 'what is the economic forecast for austria' etc.
- app cloud. Rapidly scalable, low-latency, short processes. Suitable for running a website or anything that requires huge scalability at short notice.

Mixing the 2 leads to less than optimal results.

Personally, I am expecting 'XBL gold' to be renamed as 'one cloud' (or something like that), and various features of 'app clouds' to be added. (storage, music, live data, access to large datasets etc).

In terms of 'compute cloud', I'd expect all kinds of interesting kinect usage, otherwise not much.

IMHO "cloud compute" can be seen as a higher latency, less reliable version of GPU compute - and GPU compute is something that most devs haven't got to grips with yet.

But I'd love to be wrong, as this stuff greatly interests me :).
 
Just stumbled upon this which shows Naughty Dog use Amazon's cloud computing for Uncharted.
So it turns out cloud for online isn't anything new, and if ND or whoever had wanted to use cloud processing for games this gen, I guess they could have. That's what happened with Halo, using the cloud for online aspects, but no-one is using remote computing AFAIK. Why wasn't ND using the cloud to process lightmaps, for example?

That's a question that's pretty much impossible to answer.
 
@dunbo11

I believe for Kinect upstream bandwidth is the limiting factor since at today is typically very low.
Correct me if I am wrong of course.
 
As DF pointed it out, the X1 does not seem to have special hardware that makes cloud computing easier/more efficient.

If cloud computing for games is the future, every dev/publisher can use it, interesting that ND used it already for their MP.

In this sense, MS sure is still a forerunner (lol to myself at the Halo reference) and I am really curious what new ways to ride the cloud they have to show at E3.
 
@dunbo11

I believe for Kinect upstream bandwidth is the limiting factor since at today is typically very low.
Correct me if I am wrong of course.

For the raw steam, in real-time? Almost certainly.

But there are other aspects to kinect:
- "asset import". Send the raw kinect feed for an item, and 'the compute cloud' creates a cleaned/lighting corrected & skin aligned/whatever digital representation suitable for usage in apps/games.
- character generation/mapping. Have "yourself" in-game, e.g. the mis-named "holo-presence" idea. (for example seeing your friend exercising on screen in kinect fitness, running alongside your character).

or maybe Eye-pet style things, but with the knowledge of your room layout.

Kinect is naturally compute heavy, and it should be possible to pre-process or post-process various bits - perfect for a 'compute cloud'.
 
So sum it up.

Input needs to be small.

Output needs to be small.

Cannot require immediate result.

Only thing I can think of is pre baked effects which again can be done on local hardware during loading anyways.

Yes and no. It's all relative. The input and output can be bigger if you can afford to wait (e.g. process user generated level).


Just stumbled upon this which shows Naughty Dog use Amazon's cloud computing for Uncharted.
So it turns out cloud for online isn't anything new, and if ND or whoever had wanted to use cloud processing for games this gen, I guess they could have. That's what happened with Halo, using the cloud for online aspects, but no-one is using remote computing AFAIK. Why wasn't ND using the cloud to process lightmaps, for example?

Online infrastructure has always needed a cloud platform for load balancing, fault tolerance and scaling in a unified fashion. That's why I mentioned that from user's perspective, it's not much different from existing MMOs. They would be a good case study especially as they expand to weaker devices.

But of course consoles don't typically have MMOs until nextgen ? :)
 
:p
Personally, I think it's more useful to split the cloud into it's 2 aspects:
- compute cloud. High latency, high performance, long-running processes. Suitable for questions like 'what is the meaning of life', 'what is the economic forecast for austria' etc.
- app cloud. Rapidly scalable, low-latency, short processes. Suitable for running a website or anything that requires huge scalability at short notice.

Mixing the 2 leads to less than optimal results.

Personally, I am expecting 'XBL gold' to be renamed as 'one cloud' (or something like that), and various features of 'app clouds' to be added. (storage, music, live data, access to large datasets etc).

In terms of 'compute cloud', I'd expect all kinds of interesting kinect usage, otherwise not much.

IMHO "cloud compute" can be seen as a higher latency, less reliable version of GPU compute - and GPU compute is something that most devs haven't got to grips with yet.

But I'd love to be wrong, as this stuff greatly interests me :).


(a) Live One.

:LOL::p
 
As DF pointed it out, the X1 does not seem to have special hardware that makes cloud computing easier/more efficient.

If cloud computing for games is the future, every dev/publisher can use it, interesting that ND used it already for their MP.

In this sense, MS sure is still a forerunner (lol to myself at the Halo reference) and I am really curious what new ways to ride the cloud they have to show at E3.

The VM may help make it more secure if they "open up" server component development.

For the raw steam, in real-time? Almost certainly.

But there are other aspects to kinect:
- "asset import". Send the raw kinect feed for an item, and 'the compute cloud' creates a cleaned/lighting corrected & skin aligned/whatever digital representation suitable for usage in apps/games.
- character generation/mapping. Have "yourself" in-game, e.g. the mis-named "holo-presence" idea. (for example seeing your friend exercising on screen in kinect fitness, running alongside your character).

or maybe Eye-pet style things, but with the knowledge of your room layout.

Kinect is naturally compute heavy, and it should be possible to pre-process or post-process various bits - perfect for a 'compute cloud'.

They can use the servers to help improve recognition and simulation models based on collective input (i.e., cloud source)

Most people associate cloud gaming with computational power, but for smaller devices like the iDevices, the cloud servers can run the game in a bigger memory pool undistracted by other activities. They can also aggregate data from other players on the server side. I think some turn-based phone games already do rendering on the server (refreshed every turn from crowd input). Probably save battery life too, need to test.

I find it odd that if people are so curious about cloud gaming, how come the DUST514 + EVE thing doesn't seem to gain much attention ? :) There should be a lot of interesting technical stuff there.
 
I find it odd that if people are so curious about cloud gaming, how come the DUST514 + EVE thing doesn't seem to gain much attention ? :) There should be a lot of interesting technical stuff there.


Because Dust 514 is very bad. :devilish:
Also EVE's time dilation is not exactly a universal solution since it can be applied to pretty much only turn based RPG.
 
I haven't tried it (Patch 1.0) yet. ^_^

DUST514 is in open beta. If it's like any other PS3 closed/open game betas, then you should expect it to be rough.
EVE and DUST514 share the same universe on-the-fly but DUST514 is an FPS. There is supposedly a thin client for Vita. So there should be some cloud backend to distribute game world data between the systems and clients.

Once they get the basics right and can scale profitably, they will tune and innovate further. But they are already "out there". Ideas that are impractical or don't make money won't make it there.

You can look at other MMOs for technical concepts and inspirations too.
 
No Dust 514 was officially released May 14th.
What you are playing is not a beta...the beta was even worst which is quite scary IMO.

Edit.

Earlier we were speaking of potential cloud application for Kinect and I just remembered that EverQuest has facial recognition so I wonder: could facial recognition be augmented with cloud?
 
Last edited by a moderator:
Can you clarify if you mean that decryption and decompression can be offloaded to the cloud server?
The first issue is that consoles will have some decent dedicated hardware and instructions for decryption and decompression, and the second is that data transactions over the internet should be compressed and possibly also encrypted.

well it goes like this.

1) Server side game state (encrypted/compressed)---->Client downloads game state off cloud.(like say options screen state, character select state,network settings state, multiplayer state, intro demo state etc)

2) Server decompresses and decrypts game state on client machine.

3)Client plays the decompressed/decrypted game state

4)Client requests to download new game state and the process repeats

The player only downloads just enough to play the current game sequence nothing more than that.
 
JMy point being, that the jobs for a classic cloud would have to be very specific in relation to games. BIG datasets that requires a lot of power but not a low latency response, and for a "fast response low latency cloud gaming service" it would have to be a very special cloud with very dedicated servers/cpu's to a single game.

Well well look what i found..

http://www.windowsazure.com/en-us/home/features/big-compute/

Initially we will offer two configurations (instances) for Big Compute: 8 compute cores with 60 GB of RAM and 16 compute cores with 120 GB of RAM.

Both instance configurations are delivered on systems consisting of:

Intel Sandy Bridge processors at 2.6 GHz
DDR3 1600 MHz RAM
5 x 1 TB disks
Two network connections
10 GigE network for storage and internet access
RDMA + InfiniBand (IB) 40 Gbps network for inter node communication

And between nodes...

The InfiniBand network in Windows Azure supports remote direct memory access (RDMA) communication between compute nodes. For applications, this allows memory on multiple computers to act as one.

Our RDMA network provides near bare metal performance with the elasticity and cost benefits of the cloud. Our testing shows latency of 2.1 microseconds to send a 4 byte packet across machines, and network bandwidth of over 3500 MB per sec. As a result, latency sensitive applications will scale better, with a faster time to result and lower cost. Windows Azure is the very first cloud provider to offer a virtualized InfiniBand RDMA network capability for MPI applications.

It's expensive.. but it's very fast.. perfect for a XBOX1 cloud :)
 
Right. This was the kind of thing I was imagining. There is definite potential here, again I think primarily for supporting processes for online, world-state, and processes that keep running when you are off-line / using a periphery device.

Cost is and will remain an interesting factor though. If Microsoft can make us pay for it so that the publishers can use it for free, that will increase acceptance, and there may be hope on Microsoft's side that this leads to exclusive features in games for Gold members.
 
well it goes like this.

1) Server side game state (encrypted/compressed)---->Client downloads game state off cloud.(like say options screen state, character select state,network settings state, multiplayer state, intro demo state etc)

2) Server decompresses and decrypts game state on client machine.

...
How can the server decompress the game state on the client machine?
 
well it goes like this.

1) Server side game state (encrypted/compressed)---->Client downloads game state off cloud.(like say options screen state, character select state,network settings state, multiplayer state, intro demo state etc)
2) Server decompresses and decrypts game state on client machine.
3)Client plays the decompressed/decrypted game state
4)Client requests to download new game state and the process repeats
Internet bandwidth is incredibly slow. You'll get more use out of it sending compressed data and decompressing on the client.
 
Even if you assume the Azure cloud really is so economical and useful, couldn't any competitor use it too? Another cloud platform or even Azure itself, as a client?

Whatever the market rate MS charges for Azure is... it is unlikely to be so different than their internal costs that it dramatically changes the business equation, isn't it?
 
If I were MS I would welcome anyone who wanted to run a PS4 game on Azure.

I suspect they see the future as a predominantly remote computing platform, and owning that platform is the goal.

If I were a 3rd party and I decided the cloud was the way to go and I could host my XBOne game for free, and pay to host the same code for PS4, I think I'd rather do that than try and support two different cloud platforms.

That means that publishers would be paying MS for PS4 releases, why wouldn't MS want that.

Of course it's all dependent on the cloud providing value.
 
Back
Top