Soooo, about those servers...

Status
Thread Closed: Not open for further replies.
Well, we don't play on any servers the same way we do in like WoW. My understanding is that we only submit data to either the BGS or the matchmaker; all instance connections are handled client-side; thus the hybrid client/server/p2p setup we have. The matchmaker likely couldn't keep up and thus went wonky; I doubt (though I'm no server admin) upgrading it would make our connections fair better.

This game is needing dedicated servers more and more; but Frontier went cheap with this horrid p2p rubbish. :(

IT HAS dedicated Server's. And they scale on the hour average load.
What FDev was unprepared for was teh scale of players jumping at once. Its the Transaction/Adjucation servers that are handling it. each jump makes a few ajax calls to a php backend which in turn handles the db requests which decide if players get grouped together in instances and try to balance them. You can think of it as matchmaking in a way but also more. Once players are grouped p2p comes in place for all "realtime" stuff in an instance. The problem with alot of players in an instance jumping together to a known location the Adjucation servers attempting to group the people togetehr based off wing/friends/fleet criteria against eachother player that was in the instace. This creates an ENORMOUS queryload and each request takes longer per server. If then some time out and get set back to their last savestate ( mysql-db ) and try to log in again the adjucation servers again try to group them with their friends while alot of people are still jumping or in the area. So it again may lead to a timeout -theres a point where you cant throw much more cpu's at it in time without hitting some sort of a bottleneck. The Architecture scales but it cannot handle 10% load -> 10000% load within 10 minutes. Server scale was usually over a 4hr average now its like 1hr. Its a balance act.

Developer & Serveradmin - NOT employed by Fdev btw.

Cheers!
 
Frontier doesn't own the servers, they rent them from AWS, I suspect what happen the scale of the event prevented Frontier code from scaling it server infrastructure fast enough to accomodate all of the players. They may have forgotten to account for safeguards they have almost certainly implemented to prevent frontier infrastructure from scaling randomly and out of control, say because of DoSS style attack or because of a bug in the game. I doubt it because of money.

They have admitted in the past that their code is unable to handle large groups of players jumping at once and these mass jumps have the capability to crash the game for everyone, I know there were a few large groups all attempting to jump at the same time in the same instance, at least 40 plus players.

If your online game cant handle 40 simultaneous zone transitions, you have no business making it an online only game. Because you clearly, definitively lack the tech.

Give us an offline solo version so we can keep playing the game when using it as it was designed to be used chokes the tech anyway.
 
IT HAS dedicated Server's. And they scale on the hour average load.
What FDev was unprepared for was teh scale of players jumping at once. Its the Transaction/Adjucation servers that are handling it. each jump makes a few ajax calls to a php backend which in turn handles the db requests which decide if players get grouped together in instances and try to balance them. You can think of it as matchmaking in a way but also more. Once players are grouped p2p comes in place for all "realtime" stuff in an instance. The problem with alot of players in an instance jumping together to a known location the Adjucation servers attempting to group the people togetehr based off wing/friends/fleet criteria against eachother player that was in the instace. This creates an ENORMOUS queryload and each request takes longer per server. If then some time out and get set back to their last savestate ( mysql-db ) and try to log in again the adjucation servers again try to group them with their friends while alot of people are still jumping or in the area. So it again may lead to a timeout -theres a point where you cant throw much more cpu's at it in time without hitting some sort of a bottleneck. The Architecture scales but it cannot handle 10% load -> 10000% load within 10 minutes. Server scale was usually over a 4hr average now its like 1hr. Its a balance act.

Developer & Serveradmin - NOT employed by Fdev btw.

Cheers!

To me a dedicated server is like .. WoW, or SWTOR. It's a server that we all connect to; even if our connection is rubbish, it would only affect us. In ED if your connection is rubbish and you're the host, it ruins it for everyone. So to me, that's not a dedicated server - and in most online games, dedicated servers imply servers ala WoW. For Honour, for example, was P2P and it sucked; then they introduced dedicated servers, and suddenly it's a much nicer experience.

But some helpful information nonetheless. :)
 
I strongly suspect that for it to be FD's worthwhile to ensure this doesn't happen, the playerbase in general would have to willingly accept going to a subscription model. Don't bet the rent on that happening.

I'd pay a fiver a month for proper MMO support, assuming PVP & PVe servers - or, y'know, failing that, a proper offline mode, 'cause P2P just sucks!
 
And how are dedicated servers paid for? Is that why WoW has a monthly subscription fee?
Multiple ways.
ESO is completely free to play, but those who decided to subscribe get benefits from it; like, for example, we don't pay for DLC's (expansions everyone pays for); plus we get crowns per month that lets us buy things on the in-game shop. Star Wars: The Old Republic does the same; but they limit the F2P players quite a bit (I don't agree with that at all).

ED could do something like that: 1) Optional sub with benefits, 2) support the game by buying stuff on the store, or 3) just buy the game and play it like we do now. :)


The subscription fee has been brought up before in other threads (would you pay a monthly fee for better servers - stuff like that). The response was a resounding "NO".
I would; but that's me.

Personally, I like the buy it once and play forever idea but I understand that because there's no subscription fee, these are the servers we live in.
That's why I like ESO; you can buy it once and play forever for free. Or you can sub and get access to sum stuff and don't need to buy DLC's (though if you cancel your sub you lose access to the DLC's).

If you want free rent, expect those living conditions to exist.
Also expect drunks in the hall, trash rarely gets taken out or picked up, etc.
We can always blame TJ. :)
Are you saying TJ didn't remove the trash, or that TJ is the trashy drunk? O.O
 
i find it funny how people just say p2p sucks and all and how its "the root of all ebil problems" while its actually not. if it werent p2p teh servers would have gone up in a plume of fire already. A Subscription wont help here either. Fdev started to rework teh server system in the back with the mission server. They even reworked it so each sector has its central server. But this doesent help if everyone is primairly in the same sector trying to jump as many jumps as possible.

What they need is to change the transaction system so that there's less db congestion with a multiple of players doing the "group thing". However they want to solve it the result shouldnt be xActions * xPlayers per region hammering the db. The question is how can you cache who is grouped with whom with the possibility to add new players + connections effectively without breaking things too much. Its a system that has been imüproved but they still will need to be working at it with the inclusion of fleets etc especially. But thx to that and expeditions like this they actually get data to rework a design that can handle this. You cannot stresstest Situations like this efficiently.

Its a gradual process to fix theese things and nothing you can simply do "over night" as it has to be planned correctly and identified with teh data at hand or you break more than you gain.

And no making it all "central" is no solution either or you go from a realtime game to 1-2 fps combat pretty fast once a mass like this gathers at a single point. At least like this some people stay on while the weaker ones ( net, perf, ping wise ) drop off faster. Also your just putting more work on central routers instead of each owns home router. While theese are alot more powerful its a central point again where everyone is affected by it. The STUN Proxies ED uses to connect people with unsuitable firewall setups and or without Upnp enabled are such a central thing - and if you check your connection while your connected over STUN ( Ping 400 - 800 ms ) over a correct portsetting ( ping 12 - 47 ms ) in the same instance you can clearly see teh bottleneck.

But you know what ACTUALLY helps on events like this ? ( not in case of teh adjucation server connection ) but to stay online with others ? - make your portsettings/forward on your router correct and set it in your game aswell. Sadly this needs some tech knowledge - BUT not relying on a flaky UpnP connection that may or may be not correctly handled will help you stay in thoose instances with a ton of people. If alot of people hit the instance upnp over IPv4 can cause alot of really cheap low power routers to not be able to cope with the connections and not just traffic.

Cheers!
 
Last edited:
Exactly, DW2 is a blip, catering for it specifically would not be wise. I don't doubt FD is still adjusting the server space they rent from AWS to account for the increased load since 3.3 but adjusting for the outlier of DW2 isn't feasible.

Pearl Harbor was once a blip...
 
There are no excuses for the poor server capabilities at launch, considering I received an email reminder from Frontier a week before the launch of DW2 that it was happening.
They touted it, and failed to deliver the infrastructure.
It's not like it was the end of the world either, but no one should make excuses for them.
If I was a manager/stakeholder, I'd be embarrassed.
As a player, it's just another day trying to play the game.
 
Last edited:
There are no excuses for the poor server capabilities at launch, considering I received an email reminder from Frontier a week before the launch of DW2 that it was happening.
They touted it, and failed to deliver the infrastructure.
It's not like it was the end of the world either, but no one should make excuses for them.
If I was a manager/stakeholder, I'd be embarrassed.

And from a customer perspective this is what matters.
 
I think FD responded appropriately after the first DW2 launch server gaff. Because the 2nd and 3rd launch went without the same hitch. I witnessed the second launch and it was a fun and entertaining display of the marvelous achievement of ED's model galaxy multiplayer sharing world. So they probably did what they had to do to recalibrate for the extra load whether it was renting more AWS space, or implementing more server backup code duplication or whatever.

"The launch of Distant Worlds 2 crashed Elite’s servers, but fans didn’t seem to care"
 
Last edited:
Exactly, DW2 is a blip, catering for it specifically would not be wise. I don't doubt FD is still adjusting the server space they rent from AWS to account for the increased load since 3.3 but adjusting for the outlier of DW2 isn't feasible.

It's both ironic and hilarious that you would consider 10,000+ players in DW2 a "blip", "outlier", and not "feasible".

I'd like to know what the weather is like on the planet you live on.
 
lets be honest, all players were expecting that the servers would melt, and it happened. fdev is pretty consistent with their expectations.
 
Status
Thread Closed: Not open for further replies.
Back
Top Bottom