A Visual Guide To The Server Meshing Test
HTML-код
- Опубликовано: 7 сен 2024
- How did Star Citizen's first server meshing test go? In this video I show how the shards were configured and show some footage of some very interesting server things!
🟣5% OFF PREDATOR MOUNTS LINK: www.predatormo...
(any order will support the channel)
US Parts To Buy (Amazon Affiliate Link): www.amazon.com...
UK Parts SCAN (Affiliate): tidd.ly/3TdYCJz
DISCORD: / discord
Become a member to support the channel:
/ @tenpoundfortytwo
New to Star Citizen? Use a referral code for some extra credits: robertsspacein...
Donate to directly to support the channel: paypal.me/tenp...
#StarCitizen #SpaceSim #PCGaming #Squadron42 #S42
So from what I remember from the 800 player test while I was standing in Orison when it locked up. That server was working well until around 400 and then it basically just stopped. Server fps stayed in the 20+ range. But nothing worked anymore. From my understanding of what Ahmed stated. Basically they found a bug within the servers code that was causing it to error out and it was just spamming itself with the error code, which is why it was like the world just stopped spinning and everyone and everything froze even with high server FPS still. And it was stated that they were not hitting the limits of the rep layer and it was simply the error of the server that was preventing that 800 test from working well. Which is a good thing to hear. Hopefully they get it sorted and we get another test shortly.
Great info, thank you!
Ahmed made changes for hours on the 070 server specifically and got to the point were 600 players was ok (not too different from LIVE right now) but was never able to get it to run well at 800. At 800 we got freezes, back to back crashes and the entirety of New Babbage respawned. Ahmed said it was due to other back end systems not able to utilize or account for the meshing, rather than bugs or performance issues with the meshing code itself. Things like planet rotation and gravity were the biggest offenders he said.
Interesting thanks for the info
The MVP for 4.0/SM is 4 servers, ostensibly 2 for Stanton and 2 for Pyro. Said as much in the Letter from the Chairman infographic from 2022. The reason for this is to test both the transition from server to server within a system, and from system to system, which they handle differently, as was said in the ISC about Server Meshing at the beginning of last year. You can't compare PES to SM. PES had 8 of the 12 new services needed for Dynamic Server Meshing built into it. They couldn't implement it in multiple pieces, they called it the "big bang" approach in the Post Mortem and there was no other way. That many new moving parts means exponentially more things to go wrong and exponentially greater difficulty tracking them down, not to mention the issues that only occurred at scale with live release (hence kicking it out the door from PTU, they couldn't get any more data). They reiterated all this in the most recent letter from the Chairman. 4.0 will just be Server Meshing. The Separation of the Replication Layer and Crash Recovery will already be released to live and tested. No additional new services, but some refactors of existing services (like the Mission Service, mentioned in the Monthly Reports recently) with the new services for Dynamic Server Meshing finishing development from the data we're gathering now and through/after 4.0's release. Right now we're not only testing Server Meshing, but also stress testing the Replication Layer to see where it breaks and why and what loads it can handle in order to determine how to scale it up as well with Dynamic Server Meshing with additional "Replicants."
Hearing that both the 2 and 4 server shards went surprisingly well makes me very hopeful that 4.0 will be more than just one Stanton and one Pyro server, even if the player counts aren't increased all that much and it's just the 2 server per system configuration. (I know you said don't be too hopeful, but server meshing has seemingly been working a lot better than anyone expected, whereas PES was plagued with issues from the start and didn't go as well has people expected.)
Sorry for the low energy ill voice in this one!
Get well soon brother
Don't let it happen again
@@DABOYDJEM1 I'll try!
HOW DARE YOUR WEAK MORTAL SHELL FAIL YOU?
@@gnaruto7769 😆
I'm not apart of evocati, and haven't really played SC, but I constantly keep up with updates from good YTs like you and Tomato. I especially like your content here where it's straight to the point, information is concise, and easy to digest. I don't know if this is your typical content as this is my first video of yours, but I'm definitely liking it so far. Looking forward to more.
@tenpoundfortytwo has a wealth of information on his channel. Yes, this is his typical content and I watch and subscribe for the same reasons.
I try to keep it as concise as I can! Glad you've found the channel!
hi, friend!
It seems that there is little left to be able to enjoy your tests with a more definitive version of SC in terms of local performance.
Furthermore, everything indicates that 3.23 is going to include the complete pack, in addition to the facelift that they showed us, server meshing and vulkan and the tests have begun before summer!!, these guys can still redeem themselves.
I hope they can balance, adjust and polish all this as soon as possible and we can enjoy it without limitations.
Go!!!
❤
Cool to see them getting this out there at scale so early. Hopefully these tests give them the data they need to avoid another 3.18 scenario.
Yep, no one wants a 3.18 repeat!
Every system from entitlements to the mission manager was stress tested so the amount of weird bugs we all saw is a good thing! Thank you for covering the server layouts and how the servers were split up, I was unable to get in Sunday (6+ 30k login timeouts) and Saturday all on New Babbage disappeared (including the tram I was riding in) and I fell to the surface of Microtech as the server recovered. I was then stuck in limbo for a while as the entire asset of New Babbage popped in and out of existence around me and the server kept crashing and recovering. Early days but very promising to see it working in the Tech Preview build! 🚀
Wish you touched on the rotations of the planets and how it affected the testing at the boundaries, but a great informative video. Super cool to have server meshing in public testing.
The question I had on your last video has been answered. We can defiantly see servers changing. One word - impressive. Once they polish this, we might have a game.
They showed the principles were sound, they need activities on each to synch a bit more. For instance, you are on Arcorp, you pick a box mission on Lyria, you get to Lyria, it recognised you have a mission there, you pick up your boxes and go to Walla, Walla recognises you have a mission there, you drop off your boxes. See what I mean? And THAT should be the case for EVERY TYPE OF MISSION/ ACTIVITY in the game.
Yep for a first test, I was really impressed. Like you say, making sure the missions work across servers seems like the next step!
As a long time backer that was excited to explore PO back before 3.0 when being able to run around with others in Star Citizen was capped at just a few players, the advancements from 10 to 50 t9 100 & now with potentially much more soon, I'm happy for each of the advancements and looking forward to those still to come.
CIG should try 1 server per planetary system. Include the moons with the planets. That would resolve much of the quirky shifting servers transitioning from planets to moon locations. Objects should not disappear in close proximity to the planetary bodies.
Think into the future with dynamic server meshing. Testing is to find as many issues as possible and see the current limits.
Great video. Im working on implementing server meshing for my game as well. Here's hoping I can get the same results as you.
theoretically, they need to achieve partial dynamic server meshing cause it could be that 300 players or more go into 1 moon which has only 1 server. it needs to adjust accordingly when traffic increases.
Thanks for sharing this! I wasn’t able to participate and this was a really good recap. Really appreciate it!
Thanks for the encouragement! Happy to help!
Happy to see this kind of testing happening and really involving the community.
There's always going to be people left out, I think the different "Waves" of testers works pretty well. I know people don't like that you can pay for it, but how many people are actually making meaningful IC contributions? Probably not enough to get the thousands of concurrent players they need. I think many are happy just to be able to watch this stuff whilst it gets hashed out.
The 6 server one actually worked really well if you were able to get to a moon.
I'm pretty sure the problem was that eventually too many people were in space or on a starting planet
Good info thanks- yeah from what I could tell, most people didn't make it off the planets!
This is why dynamic server meshing is the next step
@@Maverick966 yeah theoretically, they need to achieve partial dynamic server meshing cause it could be that 300 players or more go into 1 moon which has only 1 server. it needs to adjust accordingly when traffic increases.
Now, on a different note, but a related note: Pyro and all it's content, basically put each System added (each with it's own attendant content) is effectively AN EXPANSION of the game. At some the overall size of the game is going to become an issue, with regard to it's install size and footprint on the Hard Drive. I feel that it would be better to address this issue now, BEFORE Pyro rather than address it later, after the third System after Nyx drops, whether that be Magnus or Terra.
I fully expect the game to be 500Gb or more by the time it's complete. I personally will not mind.
Yep it's going to be massive eventually!
Not as much as you'd assume. Star Citizen's planets have file sizes of just a few megabytes. Sure, there will also be new assets that take up more space, but half the point of the modular approach they're taking is that they can store modular assets ONCE and use them infinitely many times.
Even the 100-system SC file size will be surprisingly small for what's in it. But yeah, in this era of primitive accumulation, there will be growth spurts.
So nice to be able to hear about Evo progress, thank you for the message to temper expectations, I do like the idea that I may have a little more interaction with other players in-game. Until now there have been so few people and so much space that most of the time seeing another person meant a firefight almost 100% of the time.
Thanks for the visual on how the servers were laid out. Its been kinda hard to understand it from other CC's.
I'm surprised it was laid out in this fashion. I didn't expect space to be its own. Or the moons to be separated from planets and grouped together.
It feels overly complicated that way. But I'm sure CIG had its reasons.
I'm more fascinated about how they can change and create the layouts so quickly.
They have gotten this far, and I'm sure they will get it working as intended.
I was there for 3.18 and remember that mess. I couldn't even log in for two months. I am a teenage whale, so money had nothing to do with as some would believe.
I just played other games in that time. Oh yeah, and touched some grass too. That was amazing, even though it was very wet in my area at the time.
No problem! Yeah I was surprised with they layout too. I was half expecting a boundary straight down the middle of the system! Was very impressive how they were able to spin up different shards with different configurations.
The testing on Saturday seemed better for me than Sunday but yeah, when it started to go unstable it cascaded all kinds of odd issues. I had inventory transactions taking 5 minutes at shops. When I was flying around early on the server transfer was smooth but them started losing my connection in space. They discovered when they did the Pyro/Stanton jump test missions would break if you changed servers so they likely have to rework abandon/disconnect logic on missions now that they can move server sessions.
Encouraging tests but needs time to cook and they seem to be moving aggressively with more devs not distracted by SQ42.
I would be very content with something like the 400 player limit server test to be able to have the best of both worlds, server meshing and server performance so we can help test within that environment.
I understand them pushing it to its limits but if they can have and maintain a stable baseline of server meshing like I mentioned perhaps dialing it down a notch that would be fantastic.
Im already impressed with what theyve achieved.
This is what we've been waiting for since 2017, I'm in awe seeing that they finally did it
Here is my thing. As of last year we heard SM would be tested this year but most of us thought it would be end of year. Many thought it wouldn't actually come for a few more years. Yet here we are. I'm just excited we're actually testing it now!
Considering the rapid pace of development since Citizencon it's hard not to get excited.
Yeah it's been pretty incredible since citizencon!
Great video. I wish other people would stop with the mental gymnastics, it did NOT work. I am hopeful they can get it sorted out for the next test and at a minimum get 200-400 players on a working STABLE server. I think it's time they start to limit their scope and get a working prototype. Think about it for a second, they are working on an MMO where the MMO component does NOT work. After all these years they are now testing the MMO component of their alpha test. That's a bit alarming after backing this game for 10+ years.
Great recap, well-explained and well-cautioned :)
I was with you on e38c (player name "Chameleon_69") and it was brutal.
How many service managers are required for 1 client, server side? How many server side server to server managers are required per shard? How much data will be copied between shards to simulate a single master shard? How Much Money will that cost Per Month?
Who will pay for those costs if we have no subscription?
All very good questions! I suppose they'll only be able to answer them after going down this road for a while. I think the cost question is a big one.
How quickly we get multiple servers per system will really come down to why why it caused problems. My biggest concern would be more database problems.
Yep let's hope its not database stuff again!
I definitely disagree about what's likely for 4.0.
The 4-server setup went reasonably well and had MUCH better performance than Live, which is pretty much NECESSARY for the more complex NPC behaviors CIG is hoping to showcase in 4.0.
As an Evo who participated in the basic single server per system ETF testing, that setup would almost certainly prove inadequate for actually elevating the player experience.
Plus Benoit is bullish af about this, and they're testing MONTHS before 4.0 PTU is even a thing. VERY different from 3.18.
Good video tho for sure. Thanks. 👍🏿
I could well be wrong! You're right, the 4 server setup was surprisingly good, so maybe they'll shoot for that. We'll see!
@@tenpoundfortytwo Tbf, I'm not quite as bullish as Benoit yet, who is rumored to be trying to debut DYNAMIC server meshing in 4.0, which while super ambitious is also pragmatic because it would save them a lot of money if they can pull that off. Moonshot within a moonshot, but we'll see. Man does know his shit. 🤷🏿♂️
my guess, and it is just a guess is that the replication layer cant cope. they must have been splitting the replication layer out previously ie you could login and the "bed" wouldnt exist on that layer.. now they cant do that. the more servers that are messed the move the replication layer gets a kicking.
also experienced unplayable but apparently they got great data and that's good enough for me
3 server config with 200 players was GOD tier
Glad to see a review of this from you
So at this rate, they may as well just get Stanton to 250/350 for a while on live, would be a nice boost while they improve it and find the sweet spot that allows them to get to 800+
Yeah that seems like one of the more likely options, but we'll have to wait and see what they do!
I think they want it all heavily tested before anything goes near the PU
Enjoying your videos Tenpoundfourtytwo!
Thanks!
Despite it being early I want them to release a post mortem for this test.
Yep I would love that, would be really great to hear what was going on behind the scenes!
Probably in the This Week in Star Citizen post, tho I hope there's a full 3.23 Retrospective post at some point.
Great video. I had similar experiences.
This is exciting. Progress is everything.
100%!
As usual, Short and Sweet, thanks
No problem!
Love this stuff, keep it up please.
Will do!
I'm extremely doubtful that we will only have 1 server in Stanton and 1 in Pyro, 2 servers in stanton and two in Pyro isn't that more complicated.
They wanna have at least 2 per system. Probably 4 per system, why? Because 1 server per system is not stable enough and too buggy
Am I confused or are you breaking the NDA showing us this? I thought Evocati were only supposed to talk about the testing?
This one was not evo- it was open to wave 2!
@@tenpoundfortytwo oh alright, makes sense then!
From a player point of view it was ok to completely unplayable ... but likely CIG got some amazing data.
So would you just be following your bounty target to whatever servers they jump to and whatnot
Once they get it working properly it should be seemless. It should work just like it does in live right now!
This is all good information and progress. We don't know exactly why the servers froze, although someone said errors on the server. I'm still pesemistic about this tech though, yes I'm sure they can get missions and quantum to work across servers which they'll need to do, but it doesn't really look like they've advanced player counts per server which is hugely dissapointing. Getting 4.0 with static server meshing and 1000 players in the shard is a pipe dream right now and something that isn't going to happen anytime soon, some might say if at all.
Blah blah blah I'm pessimistic Blah blah blah pipe dream this and pipe dream that blah blah freaking blah 😂
Well, the largest strain on the server isnt the players, but the NPCs and all the background syste,s - but when you talk about potentially 400 players on one single server, yeah kinda understandable the server buckled.
But the thing is, lets say they go for the 800 or even 1000 player per shard solution. And they do the same split on stanton, and then a "mirror" of sorts for pyro - Even with 1000 players, over 2 systems and around 12 servers, it would be significantly smoother than now. Especially in the beginning, we can expect a 40/60 or so split in playernumbers between Stanton/Pyro. For one because Pyro is just larger, but for another because pyro is newer and everyone can finally see it for themselves. And with that we'll have 400 players in Stanton (which seemed to work fairly well), and 600 in Pyro (which would imo be similar to the 400 stanton test).
Waiting for your next video on "Vulkan". Apparently it is live in the current evocati cycle. Please find someone to help u out determine how it affects the game.
(also knowing its evocati everything must be taken as a grain of salt, but yea, at least we know).
Yep really looking forward to some vulkan testing! I'll ask my evo friend, but I think the only way to really find out how it's doing will be to do lots of testing!
@@tenpoundfortytwo Agreed, hopefully we'll be able to test wave-1 soonish.
Evocati is up!@@tenpoundfortytwo
I think you are wrong with your expectation of 4.0 and CIG. The reason why they already mention 1.0 is exactly to cover themselves on the shit show we are about to experience in 4.0. they need that, they have no other solution as they cannot go commercial with full server meshing.
This is my opinion but as I said CIG is not that smart to not repeat the 3.18 experience, for them is alpha.
Yep you could be right!
@@tenpoundfortytwo I forgot to mention how valuable are your videos and your explanations. Please never stop and keep up the good work!
@SergioMerendino thanks for the kind words!
I would like to see a 800 people server but the change would be by location. I can not see the game being smooth at this stage with players from arround the world.
Fake news map at 2:30. Hurston lost its 4th moon and ArcCorp got additional 3rd. Go study map :DDD
I loved the 120 fos servers in Battlefield 4 😊
How about a server meshed cup of Liber-tea?
I came
Bro 😂
Nice, I can't wait to walk through knee deep trash...
So basically it's all down to they have not found a way to up player count, maintain the still weak stability using server meshing and we are all being told everything is great, even though no progress has actually been made.
No I did not expect CIG to know how to make this work, or expect it to work in the next three years, no matter how much gas lighting get spoon fed out
How can you call the server FPS 'relatively good' when its between 15-20? That is awful Server FPS for this game. Yes its better than single digits but its still not going to be good enough for SC. Im even doubting 30 server FPS will be good.
Ok number with no picture.
Did you even play?
@@beardedbarnstormer9577 yes i did and 15-20 fps isnt good regardless of whether whether im a number with no picture or not.... Whats my youtube account got to do with putting my point across Mr Alphanumeric name with a picture?
@@3556df44 because you're just a troll at best my dude. Server FPS when testing meshing is part of the equation. It's laughable when people like you dog shit you wouldn't understand how to build if your life depended on it. Stay fresh
This will be a carcrash. If server meshing in this form were feasible, it wouldn't take 8 years to program. In the end all we'll get will be very buggy static server meshing around planets that could've been implemented in 6 months for 1/50 of the budget. CR's fantasy world he lives in is the projects downfall..
Worked very badly.
way smaller companys made a MMO in 4 years. CIG after 11 years: a joke 🤡
server meshing is acore in every mmo
Not at this scale, or dynamic -- every MMO has loading screens and server instances. They also had to build their engine, basically from scratch, with Lumberyard as a base + all their tools. On the 11 year thing -- Read Dead Redemption 2: 12 years... GTA VI: 12 years (on release), CyberPunk 2077: 12 years (finally considered fixed and complete only this year). Point is, good games take about this length of time to make, these days, you just don't see it until about 3 years out from the planned release.
You can point to whichever MMO you want, they each have some component of star citizen done pretty well. But never seems to be "all":
Dual universe has the Super MMO (SMMO) and economy part, but it doesn't really have a player base to drive it, and it can look a bit "low-res", which isn't a bad thing.
Elite Dangerous had a good economy, had a good release, and had many things to do in a massive playworld, but there aren't as many ships and none have interiors, plus the player base is very much online solo play oriented.
Star Atlas is a "high-res" game, but tbh it's too early to tell where they're going with it? Will it be a traditional space MMO (server instances), or a SMMO (server meshing)? There's not a lot of transparency on the developers so we'll have to wait and see if this will be a rug-pull or w/e.
SC's accomplishments within the 11 years may not have been a lot, considering the first ~4-5 years was wasted on churn and increase in scope, in addition to the scrapped illfonic star marine module.
But being able to have a SMMO, "high-res" game, with fully explorable ships, and planets, and moons, in a seamless play space where people are interacting with you hasn't really been done before, and seeing progress that inches us closer to that realization, however long it is taking, is something that most of the community (who hasn't dropped hundreds of dollars on the game) finds pretty cool.
oh what a clown i have seen you in multiple comment sections go get a life
same video here as all the rest on youtube THEY PLAYING WITH AA INVISBLE BORDER INSTEAD OF TRYNA GET ALL THE PLAYERS IN ONE SPOT AND TRAVEL AS A SCHOOL OF SHIPS BUT THEY WANNA PLAY WITH A BORDER BRO I SWEAR CIG IS SO SAD RIGHT NOW ITS CRAZY
I'm extremely doubtful that we will only have 1 server in Stanton and 1 in Pyro, 2 servers in stanton and two in Pyro isn't that more complicated.
It works pretty good.
They just have to hammer out bugs and scale this up. Very promising.