Megaservers: my one big gripe
Yea the megaserver system definitely has its problems. Especially where once it decides to close a map, that map will be closed regardless of whether the population rises or not.
Yea the megaserver system definitely has its problems. Especially where once it decides to close a map, that map will be closed regardless of whether the population rises or not.
Actually, I’ve seen maps get resurrected. It’s very rare, but it does happen.
Now, it SAYS that it’s closing, but sometimes the message disappears, and you don’t get booted.
But, I completely agree that the megaserver process for meta maps is really bad.
I don’t think it’s solely a megaserver problem – I think it’s a combination of megaserver and the way most event maps run on a fixed schedule.
It actually works quite well on the Silverwastes, the only event based map without a fixed schedule. Typically when I go into the map, I’m on one that’s just about to start, giving me the choice to experience the whole thing, or use LFG to join a map that’s made some progress. The LFG is also a little less frantic, because it isn’t going to be the only opportunity to do the event for several hours.
{rant} My request is simple: allow us to select map copies to join, and to queue when it’s full, and to STOP GIVING US THE FULL ERROR WHEN IT’S ONLY HALF FULL,
JUST TO MAKE US LESS DEPENDANT ON THE COURTESY OF OTHER PEOPLE {/rant}
You only get that error if the map is really full. If you manage to get in later, it only means that someone left that map and vacated a spot.
Now, it SAYS that it’s closing, but sometimes the message disappears, and you don’t get booted.
You do get booted even if the message disappears. It’s just that it can take up to an hour. And you don’t get to see the counter then.
Remember, remember, 15th of November
I don’t think it’s solely a megaserver problem – I think it’s a combination of megaserver and the way most event maps run on a fixed schedule.
It actually works quite well on the Silverwastes, the only event based map without a fixed schedule. Typically when I go into the map, I’m on one that’s just about to start, giving me the choice to experience the whole thing, or use LFG to join a map that’s made some progress. The LFG is also a little less frantic, because it isn’t going to be the only opportunity to do the event for several hours.
But there we have the same problem, sometimes LFG has zero taxis for Silverwastes, even though I know from guildies or friends they are on an active map. meaning that without a contact on said map it’s impossible for me to join and enjoy that active map.
It’s a math problem.
- If the instance is nearly full, it soft caps, to allow for guildies & friends to join.
- If there are no other instances available, a new one is created, so you’re placed in an empty instance.
- People in an empty or near-empty instance would rather join a nearly-full one, so try to taxi, instead of trying to start their own LFG.
- If someone does start a new LFG, that instance becomes crowded, meaning new players end up in yet another newer map.
It’s not a problem ANet can ‘solve’ because it’s not a technical issue nor a design issue; it’s a human behavior reaction to the math.
ANet could design maps that have metas that are less epic and less interesting with 100 people. And they have, for the new LS episodes. Some people love those maps, because they are simple, often farmable, & and never require much time commitment.
Another strategy would be if ANet used the Silverwastes system more often, that is independent of the clock: do enough pre-events and the meta starts.
Unfortunately, in the end, each mechanic is great for some people and dull for others, so the “best” solution for the community is for ANet to have a bunch of different maps that use different methods to attract people to come back again & again. And, as it turns out, that’s pretty close to what we have in game.
I think Anet can do a lot to address this issue actually. For starters:
- If a map decides to close, give it a “return population”, where if the map reaches this population level and keeps it for at least 15 minutes (or 30, whatever) then it no longer decides to close.
- Put a substantially lower limit on how many different instances of the same map can be created at once, and only make new ones if they are all full or at least at the “return population”. I can’t count how many times I’ve been told a map is empty, and accept the transfer, only to be transferred onto another empty map that immediately gives me the warning its going to close, only to then be transferred onto another empty map, again. This absolutely should not be happening, period. If there are that few players, then that many instances shouldn’t exist in the first place for that map.
- In a very similar vein, if a player decides to leave the map, do not under any circumstances place them in another map that has already decided to close. Period. No excuses on ANet’s part for this. It should not be happening.
- Do not allow HoT maps to close in the middle of a meta.
These might not be easy to implement, but it would solve a lot of the bigger problems that players have with these megaservers.
It’s a math problem.
- If the instance is nearly full, it soft caps, to allow for guildies & friends to join.
- If there are no other instances available, a new one is created, so you’re placed in an empty instance.
- People in an empty or near-empty instance would rather join a nearly-full one, so try to taxi, instead of trying to start their own LFG.
- If someone does start a new LFG, that instance becomes crowded, meaning new players end up in yet another newer map.
It’s not a problem ANet can ‘solve’ because it’s not a technical issue nor a design issue; it’s a human behavior reaction to the math.
ANet could design maps that have metas that are less epic and less interesting with 100 people. And they have, for the new LS episodes. Some people love those maps, because they are simple, often farmable, & and never require much time commitment.
Another strategy would be if ANet used the Silverwastes system more often, that is independent of the clock: do enough pre-events and the meta starts.
Unfortunately, in the end, each mechanic is great for some people and dull for others, so the “best” solution for the community is for ANet to have a bunch of different maps that use different methods to attract people to come back again & again. And, as it turns out, that’s pretty close to what we have in game.
You make a great point. However when you are trying to do an achievement/crafting that requires one of the HoT metas, it comes down to having to deal with that frustrating issue or just not do the achievement/crafting.
Let me start by saying that I might just be lucky, but I dont have ANY of these problems. So lets have a look at said problems and see if I simply is lucky or if there is something I do to become “lucky”.
[sub]Besides that I always find myself on different instances than guildies[/sub]
Only 2 ways to ending on another map then a guildy is the following:
1) you and he are from different language servers. So someone from a EU server and someone from a ES server will have this problem.
2) the map is soft capped (or hard capped), if soft capped you can then taxi on your guildy.
The fact getting on active maps gives me such a headache.
You go to any of the four HoT maps. Once there you will always, without fail, be planted on an empty, inactive map. This has a 100% guarantee.
Not true, sry. The maps are however all fairly large and this often lead people to THINK that the map is emty. Only one where you prob will land on an emty map is DS, due to the event timer.
The next thing to do is get on an active map, which relies on other people being nice enough to have LFG queries.
Now, when it comes to the meta events, you can arrive as early as you want, it won’t matter. At some point, people are going to form the taxis to get maps full for the meta events. And here it’s not first come first serve, it’s anarchy. You join a group, and spam join on map, clicking away a hundred “the map is full”-errors (even when there’s only like 100 people in yet) and someone who joined later, may just steal your last spot away. And coming early won’t change it, because you have to wait for a taxi to appear in the LFG tool.
ok, 1) maps only tell you they are full if they are full. If you get into a map, after having gotten the “the map is full”-message, its simply due to either someone left the map, or someone dc’ed from the map. Thereby opening up a spot (this btw is the reason many org. DS maps stop taxi at the end, since they want to let anyone that dc be able to get his spot back when he reconnect)
2) you dont need to wait for a lfg to aper. Simply tag up and type something in map chat. Like “Lets get the meta rolling all, 3 more tags pls” or whatever. This lead back to the “if I dont see any tags on my map = emty map”-symdrom that many seem to suffer.
TLDR: Mega-server generally works very well. How to get into a good map? well make a map, tag up and type in map chat, be alive. Do something to org. it.
I just come from 2 “emty” maps, 1 TD and 1 AB. In both maps I simply tagged, typed in map chat to get a few more tags and made a LFG, 5 min later the map was hard capped and 5 later then that the meta was started and we made it easy in both cases. And this is how it is 99% of the times I join a map.
But surprise, someone need to tag up first, and someone need to be the first tag. So to all, tag. Join tags and it go nice and easy.
You do get booted even if the message disappears. It’s just that it can take up to an hour. And you don’t get to see the counter then.
I agree that that’s the way it’s intended to work, but I’m fairly certain I’ve played on 2-3 hours after getting one (that later went away). Once. I could be wrong, though; this game has a tendency to make you forget things for long periods of time. If my sense of reality was warped …
@Ori: yeah, despite it being substantially a math problem, there are things they can do to mitigate the impact on us. Although I disagree with some of the details in your post, I fully agree that they could do more. I’d like to see ANet offer us a few tools.
- Queuing for specific instances, like we had just after launch (when guesting wasn’t yet implemented).
- AI applied to map closures, e.g. never closing during a meta and enough new instances to match incoming players joining squads already in a full map.
- Something less random for DS. There’s no reason that people should have to join a squad and then rush to get into the same map as commander.
- Something in the LFG UI that helps us know if maps are hard capped.
- Upper right count down for maps that are closing; it should never be a surprise.
(edited by Illconceived Was Na.9781)
Back in June-Oct of last year were the last active months that the Megaserver was functioning as intended. Halloween and Christmas took huge tolls on the Legacy Player community, the new living story maps are small and easy to get the achievements/collections on even with a working life (2 weeks content max, not counting the farming of nodes).
Guild Wars 2 can persist as it is, with its Legacy Players on Hiatus. It won’t bloom again until an expansion is announced.. the Legacy Players won’t return (for any time other th an that to complete the next living story map) until an expansion is announced. Metas will continue to go undermanned, taxis will remain a ‘must’, dungeons will go rarely played (as they are abandoned features, CoE is a beautiful beast and will be missed), and Fractals will remain the blah fest that they are.
Gilded Grimoire[MAGI]. Casual Guild www.gw2magi.com
I remember the AMA a year ago, promising to fix megaservers. I hope the next expansion they care more about than this one, because new players porting to empty maps is gonna kill this game. You cannot expect players to look in LFG to find populated maps, they should be ported to them automatically.
never understood why we cant see a list of map instances and which are full and which arenot so we can pick where do go.
never understood why we cant see a list of map instances and which are full and which arenot so we can pick where do go.
Because then we could estimate the active population.
Remember, remember, 15th of November
I never understood why game companies were so hesitant to post that information.
But still the biggest changes I want to see are to never be ported to an instance that has already decided its going to close. That is infuriating and should never be happening
I never understood why game companies were so hesitant to post that information.
Because it’s sensitive business data. You don’t expect other (successful) companies to give away that kind of information.
Also, people would constantly post that the game is dead because of a seasonal drop in active players. Simply because they don’t have the full picture. People usually aren’t good with numbers, especially big ones.
I never understood why game companies were so hesitant to post that information.
You don’t expect other (successful) companies to give away that kind of information.
Telecom companies are proud to release that info; game companies are not. It’s partly a historical accident that some industries do and some don’t; it’s only partly an aspect of a well-conceived business model.
One reason why some industries don’t is that data is too-easily misunderstood. In MMOs, the number of paid customers drops off quickly after a release, there’s are different types of players (20 hrs per week for 4 weeks every 6 months vs 4 hrs per week all year), and so on. It’s easy for the media and the public to think there’s a problem if those numbers are lower after 3 months, whereas it might be quite normal for the industry.
Regardless of the reasons, ANet doesn’t release that data. But I doubt that’s why we don’t have any tools for dealing with megaservers. After all, it might be better if we actually saw that there were enough people for 6 maps worth of TD metas instead of thinking we only had enough for one? (which almost certainly happens these days, if not very frequently)
I think it’s a combination of it being more difficult than it sounds to do this with ANet’s megaserver system and ANet not realizing why it matters so much to the player experience. (They know it makes a difference; I think they underestimate the longterm impact to us.)
The megaserver is an algorithm problem and is simple for someone who is very very competent. I mean, if Jeffrey Dean from Google would tackle this megaserver issue, it would be trivial and he would have a solution in a few hours.
My problem with Megaservers is they don’t take map progression into account. Please stop kicking me off the map I just finished the SCAR lane on to put me on some kitten unprogressed map 2 minutes before Gerent. I have ley-line crystals to farm.
One of the problems is that they try to prioritize certain players. When a certain population is reached, they’ll split instances by server groupings and the old overflow logic seems to kick in, where popular servers have their own chain. They basically cap a map at 50%, where only people from the owning server(s) can get in normally, up to 75%, at which point only groups can get it.
If I show up late to a world boss for example and get into a dead map, I can usually guest to one of the popular servers to get into an active one. That’s an obvious failure of the system.
The megaserver is an algorithm problem and is simple for someone who is very very competent. I mean, if Jeffrey Dean from Google would tackle this megaserver issue, it would be trivial and he would have a solution in a few hours.
cough I can absolutely, with certainly, inform you that if you said that during an interview at Google when talking about distributed systems design, it’d put you in the bucket of “failed the interview” pretty quickly. Like, that’s pretty unrecoverable, even if you back down the system in future.
Much as you might think that there is a simple, intuitive solution to the problem, I assure you that it is neither simple nor intuitive once you get into the details designing for it — let alone implementing it effectively.
This is also exacerbated by the fact that anet operate at a scale that is at least a million times smaller than google — so the granularity of anything you can tune is really, really coarse compared to what can be done on large scale distribution systems there.
Confirmed: SlippyCheeze is Jeffrey Dean from Google.
Confirmed: SlippyCheeze is Jeffrey Dean from Google.
haha. Nah, I’m not Jeffrey Dean.