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

[Shifty Geezer]

What she's describing already exists - games run on a server. Nothing she's described wasn't conceptually possible 10+ years ago (save for the scope of the outrageous claims). Only difference now is better economics with virtual servers in the cloud. Economics are still stupidly prohibitive. "Render everything with real physics"? That's just PR sound-biting, great for securing funding from clueless venture capitalists.

What's she's not talking about is taking a locally produced game and enhancing it cost effectively with the cloud, which is what this thread was wondering about since 2013. 5 years later, we've got something going on with SoT's water. Crackdown is still yet to appear! These are seemingly the only two games using the cloud, and one is only using the cloud power in multiplayer.

 

[cheapchips]

World's Adrift, about is about to hit early access. It'll be the first biggish game running on Spacial OS. It will be an interesting test case for cloud based game worlds with persistent objects/physics.

 

[Shifty Geezer]

It's an MMO?

 

[cheapchips]

It's an MMO?

It is massively multiplayer. It's not an 'MMO' with a bit of extra physics though. Every object is physics based and persistent, as are the components of the sky boats. It also has grappling hooks.

It's still a very Bossa game, despite being online.

 

[Shifty Geezer]

Point being, they're just increasing what the servers are doing. Again, it's not local game augmentation. Perhaps we need another thread for game tech in server-based games?

 

[cheapchips]

If you keep transitioning then there's an end point where it's all server based? That how I see the thread title anyhow.

 

[Shifty Geezer]

If you keep transitioning then there's an end point where it's all server based?

In what form? If you stream the video, then it's game streaming and you can go with a thin client instead of a console. The thread title means transitioning certain aspects of a game, like physics, to the cloud. The idea started in 2006 (linked in OP) and then MS talked about doing just that this generation, bring computer power to every XB1 console.

Games with the engine running on a server and synchronising updates locally are just server based games - nothing new - and not local game augments. And as I just said, games run on a server with the video sent to players are game streaming. This thread exists for the in-between concept of server based and locally based games, of running aspects on the server to enhance what a game can do locally, especially in non-multiplayer games (single player or small coop counts).

 

[Alucardx23]

What she's describing already exists - games run on a server. Nothing she's described wasn't conceptually possible 10+ years ago (save for the scope of the outrageous claims). Only difference now is better economics with virtual servers in the cloud. Economics are still stupidly prohibitive. "Render everything with real physics"? That's just PR sound-biting, great for securing funding from clueless venture capitalists.

What's she's not talking about is taking a locally produced game and enhancing it cost effectively with the cloud, which is what this thread was wondering about since 2013. 5 years later, we've got something going on with SoT's water. Crackdown is still yet to appear! These are seemingly the only two games using the cloud, and one is only using the cloud power in multiplayer.

Crackdown 3 will be an interesting case because it will show how much better physics can be with the power of the cloud, because you will also have an offline mode that doesn't have it. This will spread the word and increase awareness about cloud processing. Is there also any reason why the sea simulation in SoT could not be shared in a single player game?

 

[milk]

Crackdown 3 will be an interesting case because it will show how much better physics can be with the power of the cloud,

...assuming it works well.

 

[JPT]

I am still beating the latency drum

https://blog.packet-foo.com/2014/09/how-millisecond-delays-may-kill-database-performance/#more-642

And

http://www.lovemytool.com/blog/2013/08/its-the-speed-of-light-dudeby-dc-palter.html

It's possible to engineering away some parts, but some things will not be possible to do anything about.
I would worry about unstable links like WiFi that can suddenly spike your delay, due to intermittent/random noise. And basically all traffic going over your WiFi infrastructure is classified as Best Effort and it will get stuck in aggregation queues etc. Then add some packet drops on aggregated frames and you got even more delay. And it's unpredictable, unless you buffer for quite a while, i am guessing 500ms or more?

//edit
Bunch of spelling mistakes, thanks to my mobile keyboard.

 

[Shifty Geezer]

Crackdown 3 will be an interesting case because it will show how much better physics can be with the power of the cloud, because you will also have an offline mode that doesn't have it.

Except the fact the single player mode doesn't have it shows probably it's economically not viable, which is one of the major arguments against. If it is economically viable (and it's more cost effective for MS than other console company because they already have unused cloud capability), then why isn't the same code in the single player? Or is it to keep the single player experience uniform offline (another reason why cloud augments may not work)?

This will spread the word and increase awareness about cloud processing. Is there also any reason why the sea simulation in SoT could not be shared in a single player game?

If the sea is simulated with 200 ships on it, and that info is used for 200 separate players who can't see the other ships, some of which may be in the same space, it wouldn't work. You'd have dents in the sea where the invisible hulls of ships in another parallel universe are affecting the sea's form. If it's just a height-field that ignores the presence of other ships, they could simulate the sea on one server and stream it to every player including multiplayer.

 

[Alucardx23]

...assuming it works well.

Sure, I'm assuming it will work, but it might as well fail miserably. Let's see what happens...

 

[Alucardx23]

Except the fact the single player mode doesn't have it shows probably it's economically not viable, which is one of the major arguments against. If it is economically viable (and it's more cost effective for MS than other console company because they already have unused cloud capability), then why isn't the same code in the single player? Or is it to keep the single player experience uniform offline (another reason why cloud augments may not work)?
If the sea is simulated with 200 ships on it, and that info is used for 200 separate players who can't see the other ships, some of which may be in the same space, it wouldn't work. You'd have dents in the sea where the invisible hulls of ships in another parallel universe are affecting the sea's form. If it's just a height-field that ignores the presence of other ships, they could simulate the sea on one server and stream it to every player including multiplayer.

I found these comments from the Crackdown 3 design director. At least publicly they are not saying it was because of economic reasons..

"Wilson explained that Sumo decided against putting the dedicated server-based destruction in the campaign, which can be played cooperatively by up to four people, for a couple of reasons. One, so the game could be played offline. And two, because the story is about saving the city of New Providence, not destroying it."

"What I didn't want to do is put it into the campaign because it would have meant the entire game would have required a constant, high-quality internet connection," Wilson said.

"I wanted people to be able to play it when they went away on holiday.

"Also, the destruction didn't work narratively. You're meant to be saving the city. It was a much longer game experience.

https://www.google.com.do/amp/s/www...confusion-surrounding-crackdown-3-destruction

On regards to SoT I would like to understand better how the simulation/server works. So if the server is broadcasting the sea simulation and every client is affecting that simulation, can't the one playing the game single player have access to the base simulation without the effect of the other boats?

 

[Shifty Geezer]

"I wanted people to be able to play it when they went away on holiday.

That's one of the concerns raised over the concept of augments. That doesn't mean economics doesn't come into it though. If they did want local destruction, then economics may still have been prohibitive.

On regards to SoT I would like to understand better how the simulation/server works.

We don't know.

So if the server is broadcasting the sea simulation and every client is affecting that simulation, can't the one playing the game single player have access to the base simulation without the effect of the other boats?

How do you model the sea with both hundreds of ships on it and only one ship? Consider a simulation with 2 ships. The waves deform realistically around the ships and data is included to show what shaders to use, to add spray and whatnot...



If you pass that same water simulation to all single players, player 1 will see the waves deforming and spraying where ship 2 is in the simulation, without ship 2 being present in player 1's game.



What you'd need in that case is a hybrid solution where the general wave simulation is computed in the cloud, and local simulation is performed on top of that for every vessel in the local game.

 

[Alucardx23]

That's one of the concerns raised over the concept of augments. That doesn't mean economics doesn't come into it though. If they did want local destruction, then economics may still have been prohibitive.

We don't know.
How do you model the sea with both hundreds of ships on it and only one ship? Consider a simulation with 2 ships. The waves deform realistically around the ships and data is included to show what shaders to use, to add spray and whatnot...

View attachment 2533

If you pass that same water simulation to all single players, player 1 will see the waves deforming and spraying where ship 2 is in the simulation, without ship 2 being present in player 1's game.

View attachment 2534

What you'd need in that case is a hybrid solution where the general wave simulation is computed in the cloud, and local simulation is performed on top of that for every vessel in the local game.

Sorry, I didn't explain what I meant with base simulation. Basically is the same thing you refer to when you say "general wave simulation". As you say a hybrid solution should work where the local xbox is receiving the general wave simulation and only calculating the effect of your boat in the sea, the rest of the simulation is handled by the server and it looks the same for every other player. In the case of Crackdown 3 I'm trying to visualice what would be the difference between a multiplayer match and a single player one when it comes to cost. Crackdown 3 will allow for 10 players to play together online. Let's say that each one of the 10 players shoots at a different building in the world, all of the actions from the 10 players must be sent to the server so the physics can be calculated and then synced with the rest of the players. What would be the difference between this scenario and having only one player shooting at a building in a single player session? The way I see it, the same processing power required to calculate the physics of single player actions in a multiplayer match is the same as the same player in a single player session. One thing that I have to mention is that I do agree that for this to happen you must pay an Xbox Live Gold subscription when playing a single player/cloud enhanced game.

 

[Malo]

How do you model the sea with both hundreds of ships on it and only one ship? Consider a simulation with 2 ships. The waves deform realistically around the ships and data is included to show what shaders to use, to add spray and whatnot...

SoT's implementation seems to be far more basic than the impression it gives. There is no interaction for the water for anything, it's a simulation of water and that's where it ends. They an achieve what they have because the ships react locally to the water and can be synchronized across multiple clients because the water is identical across all clients and doesn't change. As you're mentioning, being able to track water changes isn't really feasible across clients at a decent tick rate so we end up with what we have, the hybrid:

What you'd need in that case is a hybrid solution where the general wave simulation is computed in the cloud, and local simulation is performed on top of that for every vessel in the local game.

 

[Shifty Geezer]

Let's say that each one of the 10 players shoots at a different building in the world, all of the actions from the 10 players must be sent to the server so the physics can be calculated and then synced with the rest of the players. What would be the difference between this scenario and having only one player shooting at a building in a single player session? The way I see it, the same processing power required to calculate the physics of single player actions in a multiplayer match is the same as the same player in a single player session.

Best case, one server can simulate the city and the destruction 10 players create, at one tenth the cost of the same server requirements for one player. The real-time nature of the game means you could add on more server power when things got busier, and would have to allocate enough up front. Worst case is as you suggest, that you need x amount of power for one player, and 10x amount of power for ten players, but that's not how servers normally operate and I can't see that being an issue.

Thus the economics I think are effectively $n thousand in servers to support m thousand players, which would only support m hundred players if playing solo.

 

[Alucardx23]

Best case, one server can simulate the city and the destruction 10 players create, at one tenth the cost of the same server requirements for one player. The real-time nature of the game means you could add on more server power when things got busier, and would have to allocate enough up front. Worst case is as you suggest, that you need x amount of power for one player, and 10x amount of power for ten players, but that's not how servers normally operate and I can't see that being an issue.

Thus the economics I think are effectively $n thousand in servers to support m thousand players, which would only support m hundred players if playing solo.

I was under the impression that X amount of power was reserved on the server side depending on the simulation/amount of players, so it didn't matter if the players were together or not..... I need to educate myself better on this.

 

[AzBat]

Why would they model the global water shared amongst other ships during this theorized single player game? If you're playing single player then there wouldn't need to be any ships. Your ship would be the only one since there are no AI ships. @Malo is right about the water. There is no water interaction. The boats react to the water & it's not deformed. Plus, it's only shared on a per server basis which has at best 8 ships per server or about 32 players. A single player experience would have to allocate a different sea model for each individual single player game. Like Shifty said it's probably not economical to do so. Hence the multiplayer-only nature of the game.

Tommy McClain

 

[Scott_Arm]

Well, "The Cloud" is software and services online so ... servers, which we've had forever. But if you're looking for examples of games that are offloading more processing server side, then Battlefield 5 is probably going to be a game to watch. It's going to have a more sophisticated destruction engine, and more determinism, so all clients will see the same thing. Data rates for playing online games are increasing, and battlefield 5 will probably not be an exception. Is it a revolution? No. More like an evolution of what we've had for yours. But there's no doubt that in online gaming the server is becoming more and more important.