Jump to content

Recommended Posts

4 hours ago, Lethys said:

Just watch the video i linked. He himself said that you won't be able to play with a thousand players. That's an instanced Server and no single shard 

Yep, that has always been the plan.  Not really sure how people missed that... 

 

Hopefully not too many people are under that impression, or they will be sourly disappointed.  I suppose the size of ships like the Retribution gives credence to those misconceptions.  

 

Shards and servers is a good way to put it.  I imagine it will work a lot like SpatialOS... but their own proprietary technology.  We will have to see though

Link to comment
Share on other sites

I could see the NQ server tech being much more like a SpatialOS competitor, in fact I could see Novaquark as a company actually being founded  around  building and marketing a SpatialOS like product on top of which they are building DU as a showcase product with the OS/architecture underneath being the actual asset. It would 'explain' the interest of VC investment much more than any interest for the game itself from that corner.

Link to comment
Share on other sites

18 hours ago, Hades said:

Yep, that has always been the plan.  Not really sure how people missed that... 

 

Hopefully not too many people are under that impression, or they will be sourly disappointed.  I suppose the size of ships like the Retribution gives credence to those misconceptions.  

 

Shards and servers is a good way to put it.  I imagine it will work a lot like SpatialOS... but their own proprietary technology.  We will have to see though

I don't think it can. If you browse through how SpatialOS works, you get a rough idea what it requires. Do note, SpatialOS itself has been being developed for a good while now with hundreds of millions of investment VC money into Improbable. Even then, their flagship title, World's Adrift has been no picnic of easy development (though it is quite impressive; but you notice them and Seed both go for really 'EASY' graphics (good choice!)). Both CSSU and SpatialOS are able to leverage cloud networking to DYNAMICALLY SCALE over a FLEXIBLE AREA, first of all. But the way they both seem to work as far as I can make out on scraps of info (CSSU) appears to be quite a different approach (to the same problem with a similar solution). This difference is quite simple to explain, if I understand it right: SpatialOS is able to convert the game info into "elements": Anything that's usually an object with properties or other systems eg physics even or AI etc. Basically this abstracted system can simply add servers or workers (I think they're called in SpatialOS language) to deal with more elements in a given area. I think the magic is to use this info to simulate the game state then feed the final version (resolve failures more effectively and capture a shared picture for everyone) to all the clients and all the while dynamically accounting for everything. With CSSU, instead of "workers", they use "actors" which I believe again leverage cloud servers to deal with not more elements, but more pockets of space up to 8m but adaptable in dimension DYNAMICALLY iirc in game space per actor. Now the preceding may be deeply flawed, but I have followed it to that point, again I don't know what CSSU is doing, but I guess what makes CSSU impressive is that these pockets of space for some reason I am unaware of (is there high level of redundancy?), manage to provide impressive results with respect to the load and to performance. Perhaps it's because the space is not changed when the scaling happens (it's far beyond my understanding)? If you look at CSSU and SpatialOS in videos, there's some sort of "slight sluggishness" on the load, so both solutions are demanding and come with performance costs, but CSSU seems less sluggish than what I have seen with SpatialOS, even.

 

Ok, I really hope some of the above is accurate!

 

But looking at the above, then wondering if EITHER solution could be used by CIG for SC? I cannot see it being at all possible. SC in CryEngine/Lumberyard can't use API of SpatialOS. However CIG has developed their Lumberyard/StarEngine, it's still using the old netcode of CryEngine, the way the engine works despite all the meticulous changes will still be that performance hog that CryEngine is RENOWNED for being stuck in the guts of the technology (benchmark tests iirc in computer shopper years ago?). SC assets were designed to be "high fidelity", complex data on top of the above (they spent ages "refactoring" before the current attempts to cull network data (or what the client is reporting/seeing) via OCS (only client so far) BC. The results of this: Problems escalate from the improved fps when: Large assets are loaded eg cities, AI NPC routines are high, gameplay dynamics are high eg combat.

 

They can't use SpatialOS as Lumberyard has a deal with AWS, SpatialOS with Google. Besides NQ, is working full time on developing their CSSU solution practically via DU. Secondly as already said, CSSU is a priori solution. You can't dump it onto a game. It's whole paradigm is simplifying the data to be able to handle it via the network (as SpatialOS does).

 

So is CIG going to come up with their own comparable solution? No. They don't have the expertise, it won't fit their engine, it won't fit their game, the focus on the game doesn't necessarily benefit from the solution in the first place for the limitations it imposes so it does not even make sense from the design pov either. So far only SPECIALIST EXPERTISE has come up with these 2 solutions. Another MMO in development by City State Entertainment, Camelot Unchained have also attempted their own solution to scale numbers in a dense space, and that is probably a good reality check on what limitations are imposed. Honestly I will never know why an Open World with popped Combat Arena instances is not used by CIG...

 

Coming back to SC, and "server meshing": They intend to create linked instances (which is what they already do) via (assuming this is the plan) providing more instances per star system than currently work. So far the instances are linked via hidden loading screens (apparently the most logical explanation I've seen as I don't play the game). What Server Meshing is intended to do is link more instances and again "hide" the loading between each. Now I don't believe it is possible to DYNAMICALLY shift the zones in SC. I don't know what is possible in fact. But the best I would guess is that they can add static zones and thereby have a higher total population per star system (of which they only have 2 incomplete so far and will have probably 5 complete at MVP official launch). Then for all reasons there is an upper limit on player numbers for POI in particular and maybe therefore for star systems too? The "handing off" between linked instances, it just seems fraught with difficulties and complexities with little upside, going by what they have now. Some say that CIG have solved everything to date, but their current game is very tech demo and incomplete so that is incorrect. But what to expect: They must hit limitations at some point irrespective of the funding bonanza. I suspect the limited networking "MMO" is it. This is all a far cry (that was not a pun) from the 3 other games mentioned here: World's Adrift, Dual Universe and Camelot Unchained.

 

Please correct me, super tech guys and gals if you spot mistakes which must surely exist in the above: It'll help everyone's understanding to advance a more complete picture. Apologies for the wall of text.

Link to comment
Share on other sites

5 hours ago, Borb_1 said:

But looking at the above, then wondering if EITHER solution could be used by CIG for SC? I cannot see it being at all possible. SC in CryEngine/Lumberyard can't use API of SpatialOS.

Why?

5 hours ago, Borb_1 said:

They can't use SpatialOS as Lumberyard has a deal with AWS, SpatialOS with Google. Besides NQ, is working full time on developing their CSSU solution practically via DU.

Does the Lumberyard user licence explicitly forbiddes the use of SpartialOS?

 

 

CIG now has their container streaming, which could potentially increase the amount of players in the world, because they could run each container on one server, or be calculated by one.

Also this helps with performance for players, because if there are players inside a ship, then their information only needs to be replicated to players who are also inside that ship.

They still will have the problem with a lot of players in one location (container).

Link to comment
Share on other sites

1 hour ago, Takao said:

Why?

Does the Lumberyard user licence explicitly forbiddes the use of SpartialOS?

 

 

CIG now has their container streaming, which could potentially increase the amount of players in the world, because they could run each container on one server, or be calculated by one.

Also this helps with performance for players, because if there are players inside a ship, then their information only needs to be replicated to players who are also inside that ship.

They still will have the problem with a lot of players in one location (container).

It could be made to work using the SDK on "any engine" but CIG have their own highly modified StarEngine/Lumberyard fork which I am guessing complicates an already complicated development: For example one of the devs working on CU, summed up their impressions with respect to engines in general, as such:

 

Quote

Every library and engine has what Ive called a grain, a set of easy things to add/subtract and a set of difficult things. The more complete the engine, the more opinionated is the grain. If an engine tries to have too many easy things, you start paying more and more of a performance cost. Now granted for a lot of projects, as long as youre not pushing any boundaries this can be fine and you can work within the grain of the engine. 

I thought this was worth considering with respect to what CIG have been doing: IE CryEngine/Lumberyward is an "Opinionated" engine and CIG are "pushing the boundaries" hence the numerous performance problems - already, then throw in the actual development which requires a substantial conversion for the sdk to be working with the chosen engine (untested) and also host in live environments - finally. There is now since I last looked at SpatialOS, 2/3yrs or more ago, a game called Mavericks attempting a 1,000 ppl number for a FPS game. But as they point out, they designed from the ground up along the same paradigm of SpatialOS to attempt such an outcome. Secondly that's just the claim. They have to actually develop it successfully with all their advantages already in place eg going with the grain of the engine, a relatively simple game design. An example might be that the client loads a ton of things in SC (and the devs have been culling the data the client is seeing to increase performance); whereas any world made in say SpatialOS, the client only sees what it is interested in, the rest of the world is not on the server which is hosting the client. At least that's one thing I remember from the SpatialOS video showing dynamic server handling (increasing nodes around more players eg).

 

Lumberyard license must use AWS. Not something I took much notice of but seems a non-starter even before all the above. Yes you are probably right about the container streaming bumping up numbers.

Link to comment
Share on other sites

Given how the decent ships cost hundreds and up to thousands of real world money and how the game has progressed little from what was originally promised back in 2011. Star Citizen just seems like a glorified cash grab that continues to dangle the carrot in front of people with small new features to keep them interested or draw in more people. I get why some people are so hopeful for SC and that is because they have sank incredible amounts of personal money into it.  The game has profited hundreds of millions from people and it still provides so little in return. But hey, as soon as you give them that next $100 the game will be better. Then the next $100 will actually get the game done and it remains a never ending loop of just a little more. I hope people enjoy paying for those expensive cars and housing the creators of SC now have. Because surely those items will make the game better, and hey just a little more help in donations will get the game done, they promise. ?

Link to comment
Share on other sites

There's some more networking information from back in 2017 by Clive Johnson, their (iirc) Chief Networking Engineer (nice guy too) that zeroes in on CIG's networking solution: "Server Meshing". Tbh I remember reading this, it feels years ago now, but coming back to it, I'm impressed with how succinct he's managed to convey the entire idea in such a short but useful description. Worth a read:-

 

Quote

In a single server instance we can currently have up to 40 players in Area18 or 24 players in Crusader. Matchmaking tries to put you in the same instance as your friends, but beyond that it is luck of the draw which instance you will end up in. However [atH0wland] is correct in that our goal is that eventually everyone will be in the same instance.
There quite a few engineering hurdles we need to overcome before this can happen. Server performance needs to improve a lot, so there are several tasks to address this that are either currently underway or in the schedule. This will only get us so far though, and won't be enough to fill a solar system with players and NPCs. To go further we are going to have to connect multiple servers together in something we're calling a "server mesh." Each server will take on the processing load for a region of space, and these regions will adjust their boundaries to best balance that load with their neighbors. You will be able to see (and fire) across the boundary from one server to another, and, as you fly through space, will move seamlessly from one server to another. We will also be able to dynamically add and remove servers to suit the current level of demand. This technology will allow us to scale almost without limit while keeping everyone in the same instance.
The problem we still need to figure out is how to handle everyone heading to the same place at the same time. I'm not sure there's an engineering solution to that one, so it may require some game mechanic to prevent too many players congregating in the same place.
TL;DR - yes, once all the pieces are in place and the kinks have been worked out, you'll be able to stalk your prey, and should always be in the same instance.

If you follow the highlighted bits I think it describes what will happen (wide space).

Link to comment
Share on other sites

  • 2 weeks later...

It's not so matter of CR not wanting to make it with thousands of people its simply the way he's making it such a thing would be impossible with current technology, your likely to get around 150 maybe stretching to 250 per instance eventually basically Planetside 2 is the sorta peak of what you could expect, however instances will be sorta dynamic, and cover a solar system, the greater the no of people you put on screen the less fidelity you can get out of them, Star Citizen has true fps like handling, in order to track everyone with that level of fidelity you just cant fit in to many real players.

 

 If you ever played games like Darkfall that and that only had rpg mechanics with a small amount of fps in magic and archery, it turned into a lag fest when more than a couple of hundred people showed up with slide show frame-rates when it reached thousands, and eve which at its heart is a powerful visualisation layered on top of a complex spreadsheet, still only achieves large battles using "time dilation" as in deliberate lag to give the servers the time to actually process everything, which wouldn't work to well in a game with first person, put simply no game as of yet has managed to pull of large scale battles without unacceptable levels of lag, even much simply top down rpg types still basically have a couple of hundred on screen limit.

 

There is a possibilty that Camelot unchained which has an engine built from scratch for this purpose may be able to get to thousands on screen, but it does this by setting the graphical fidelity for each character to something similar to what a game 20 years ago could pull off, its basically using the power or moore's law to make a game of yesteryear run with allot of peeps on screen, perhaps in another 20 years such a thing might be possible with SC level graphics also due to said law... but people aint gonna want a 26 year dev cycle :D

 

They will do the best they can to make performance capable servers and stretch the pop without killing perf in final game, its the best anyone can hope for, for now.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...