dynamicbob
Banned
thats probably why background sim and CG updates are handled hourly instead of in real time.
Servers cant keep up?
- Thread starter BluStar
- Start date
Yup. When I start to play in the evenings (my time), the game is smooth cause it's early morning in the UK. But as it gets later in the night the server response really starts to suck. Last night I was stuck dropping out of SC into a CZ. I had to kill my client and restart it.
For the last few nights when I want to save and exit to desktop, the screen just hangs on a black screen until I kill the client.
I have also been getting some "Mission server unavailable" errors.
For the last few nights when I want to save and exit to desktop, the screen just hangs on a black screen until I kill the client.
I have also been getting some "Mission server unavailable" errors.
It'd due to the threat of instant ship transfers!
Ok more seriously, it used to only be an occasional issue, and mainly when in a wing. Now it seems I can be sat in a starport and go make a coffee, and when I come back i have been kicked. NM the time it takes sometimes to enter/exit sc or Jump.
Is all a little worrying, I am hoping it is due to a huge jump in new people/old players coming back with the prospect of 2.2.
Ok more seriously, it used to only be an occasional issue, and mainly when in a wing. Now it seems I can be sat in a starport and go make a coffee, and when I come back i have been kicked. NM the time it takes sometimes to enter/exit sc or Jump.
Is all a little worrying, I am hoping it is due to a huge jump in new people/old players coming back with the prospect of 2.2.
dynamicbob
Banned
If Fdev are having issues paying for the services i dont mind paying to name new colonies, paint jobs and the like but give us some variance or a procedural paint job system
Oh I imagine they can afford the AWS bill, it's just that they know that AWS is hideously bad for realtime work like well, what they're trying to make it achieve with Elite Dangerous. AWS like most cloud services is -great- for generating masses of data or databases that you can readily search through and pull data down from, if you want to query something, or you want to run a shop front, AWS is fantastic. The problems start happening when you've got a steady stream of *two way* communication.
Case in point was the Simcity reboot. Everything was tickety-boo for the reviewers because they were running -well- within the margin for error that AWS can handle for realtime loads (protip: their realtime capacity is usually around 10% of their batch capacity, so if you imagine that you have around 50,000 concurrent users in spikes on a game, then you have to budget for a maximum load for -half a million- users to give yourself the appropriate room) but the moment the real users started whacking on AWS with sticks? Well that didn't end well at all did it?
Herein lies the problem for Frontier, things like the Galaxy simulation and the BGS are basically big databases, in the BGS's case, not even a very complex one by my reckoning (I've replicated what the BGS does in Microsoft *Access* just for giggles, the Galaxy sim however is a very complicated work of art). These are loads well suited to things like AWS because you're creating big fat databases which can be mined for information. Where things fall down is when you've got stuff like the Transaction server and similar which also are handled by AWS, and that's -not- work which AWS is much cop at. That's when the tractor analogy comes into play, because annoyingly you can't really do half of the stuff in the cloud and half of it on local servers, not without some really, really nifty netcode to handle stuff like servers going up and down and the like, not to mention latency and the whole "keeping things in order between databases" stuff.
The fix would be to localise the lot and have FDev do all the hosting, but that's really costly, like *REALLY* costly. So, you'll have to live with the server outages.
I guess that would explain why simple functions sometimes takes ages to execute like opening the galaxy map, selecting station services when dock, or just a general lack of responsiveness at times.
Now you have me curious- why would FD not be able to use AWS to provide a service AWS is actually designed for, ie, the quick up- and downspinning of instances depending on demand? I understand from your posts that you have a background in DB engineering (or use) but I'd hazard a guess that AWS as a tool is exactly right for this type of use case short of building your own infrastructure. I'm not saying FD got this right, or that their load prediction algos need a bit of tweaking, etc etc, but I'm honestly curious as to why you think AWS is the wrong tool for this job. PM is fine if this gets off topic.
If you want the nitty gritty and are interested enough here is a presentation on how it all hangs together that the Frontier server guys did - just over 50 minutes long.
http://www.youtube.com/watch?v=EvJPyjmfdz0&sns=em
Mark.
http://www.youtube.com/watch?v=EvJPyjmfdz0&sns=em
Mark.
Rep'd, nail hit squarely on the head. Until they upgrade, this will never improve.
No , the galaxy is procedurally generated including stations, planet surfaces, bases, names etc... It's not about a big database, problem is that it's all generated including missions etc... Server probably sends the seed to client.
When you move,shoot, activate FSD, Modify ships, jump to another system or land, it all synch with server, database only contains names the server uses to name npcs, stations and many stars, data about handcrafted systems like SOL and a few other stars probably, your ships data, position, landmarks, references to asset client has to load, the rest is generated on the fly in a procedural way so everything is the same every time you visit a place including orbits, planets' day/night, this is why the server is having so many issues, they need a better server to handle all of this, better connectivity to external networks, higher RAM and better processor ...
I think you want to do a bit of reading and fact-checking as the above has serious room for improvement
What do you mean? isn't the galaxy procedurally generated ??
- - - - - Additional Content Posted / Auto Merge - - - - -
If you want to know how ED works, maybe you can read this http://elite-dangerous.wikia.com/wiki/Procedural_Generation
Isn't the background sim a completely separate thing to the procedural stuff though. I think this is what the other comment was about and this is the reason why the networking is problematic.
Sooooooo, enlighten us all Cmdr. It would be good for the whole forum to know so it might cut down the ignorance to how things work and cut down on these post's (Maybe).
Interested!
No, it's explained in the official wikia , background simulation such as planets' orbits, rotation, inclination, masses are all procedurally generated, I think they use UTC to keep everyone synched, everything sees the same things cause it's all synched through procedural generation + UTC, probably client just decides what to render based on those info, the algorithm makes sure it's all perfectly synched with UTC, I think that's why there is a UTC in place and hour is universal and not depending on where you are
- - - - - Additional Content Posted / Auto Merge - - - - -
Don't feed the troll LOL
Last edited:
It's really all been said in this thread. Read ThatFuzzyTiger's comments on AWS (while I disagree on some finer points his comments on AWS as a platform are well thought-out and made), watch markdiss' video and read up on other comments about how the networking is implemented.
As steve said, 'the background sim a completely separate thing to the procedural stuff though'. 'When you move,shoot, activate FSD, Modify ships, jump to another system or land, it all synch with server, database only contains names the server uses to name npcs, stations and many stars...' is just cringeworthy and betrays either a language issue (in that case my apologies) or a serious lack of understanding of networking, how the environment is set up and how the mechanics work in general.
That's where I'll leave it at.
What's wrong ? So you mean that when you move or modify your ship buying a module or jump to another system you have no synch with server ? You don't notify server you moved to next system? That's laughable you think that since it is explained in detail on the wiki and according to technical data it is exactly what's going on, also stuff like commodities market rely on servers, the whole economy and Powerplay do as well, even local factions' standings and events, war zones...
I think the lag has nothing to do with the network but how overloaded EDServers are
I think we can all agree that something needs to be done otherwise it really will be the fly in the ointment for years to come !
I imagine what the game would be like if the networking worked smoothly. Players can all connect properly, wing up, no rubber banding, orbital cruise drops are instant instead of waiting an age and stopping the progression toward the planet etc etc. Just imagine...wouldn't that be cool. Unfortunately the basic framework for the network is never going to be able to accommodate this. Maybe in 10-20 years when global communications are quantum based !
The actual amount of network traffic is rather small, this is (very rarely) a question of getting data to and from the server. The delays on (for example) existing supercruise/orbital cruise happen primarily in the servers themselves - even in solo where there is no matchmaking with other players required, you sit there for an awfully long time waiting for the exit animation to conclude.
Frontier is not responsible for the everyone elses routers, cabling and ISP screw ups on the Internet. Nor can it be held accountable for bad weather, sunspots, electrical issues, other traffic demands (You may not realize it, but there are several others on the Internet not playing E: D) or the poorly configured, out of spec and ill-used equipment in peoples homes. If you are having an issue, open a ticket and send @Frontier_Help a tweet for assistance.
Frontier Server status is listed here: https://community.elitedangerous.com/
Or check here when you have issues: http://www.internettrafficreport.com/
"Cheaped out on the server solution and are using AWS" - Please, give me your authoritative opinion on what global network is "better" than AWS. Who is running a more cost effective and better rated service that exceeds AWS's reliability? I'll be having drinks with at least three CTO's/CIO's this evening who are currently (and happily) running AWS for multi-million dollar companies. Both they, and I, would love to know of a better solution.
Last edited: