there have been ongoing backlog problems with some servers in the past weeks. According to Martin these problems are not hardware related, and I have no reason to doubt his words. That leaves
software problems like data base performance or access speed,
great number of possible routes to fly from A to B (read: too many planes flying around),
great number of airports,
great number of players.
I understand that the AS team would not like to reduce the number of players - or the number of airlines - as these are needed to make a profit. I also understand the team would not like to simplify the game. It’s complexity makes it attractive and different from other games. But waiting 20 seconds before the flight scheduling page shows the changed flight, or waiting 1 minute before my statistics page opens… that is internet from the early nineties
But two things can be changed without making the game less fun to play.
Ground connections between airports. Sometimes small airports that are not being served by any airline take a long time to calculate their demand. I mean, a one green bar airport is small. And 15 seconds is a lot of time if you know that there are only 20 seconds between the passenger demand calculation of airports. The reason for this is the ground network. Passengers from small unserved airports go to other airports to reach their destination. I suggest to reduce the radius of ground networks. Perhaps to 100 km or so. I would not even mind if ground networks were completely disabled until the system performs better. If I want the passengers from some small airport, I shall simply have to fly to that airport.
Another possiblitity is to reduce the total number of passengers. Less passengers would result in less flights, and less server loads. The only difference would be that instead of competing to fill the 10th daily flight between two airports, we would be competing to fill the 8th daily flight. 20% less passengers might hopefully result in 10% less planes flying around (on some routes people would use smaller planes without reducing the number of flights).
Anyway, this sort of reduction would not make the game less complicated, less challenging or less fun. But it would improve game performance.
Ground Connections: You’re absolutely right. They sure are a cause of higher load for the servers. I can’t tell how much exactly, but it would certainly help. Under one condition: No adjustments to the transfer times like it’s the case on Devau. This game world clearly demonstrates that both changes combined can even make things worse. Ultimately though I want to get rid of the ground networks because they are a bad model for something that we wanted to have in the game but the system can’t handle very well: Competition between airports or put differently, regional demand instead of “point-based demand”. Since demand in AS is essentially the latter, ground networks are way to spread it out at least a little bit. I’m thinking a lot about an alternative model but have too many other things to work on right now.
Overall passenger numbers: In fact the demand-level should be somewhat dependent on the current amount of players and airlines. There are two major catches to that: Directly binding it to the amount of players would probably make competition a tiny bit more boring because every player essentially get’s his share by just signing up and tweaking demand always causes issues with thin relations that only see very view passengers a day (so adjusting traffic just a little bit might have severe effect on small airports while it’s only noticable as a regular traffic reduction on large airports).
Both options have the common potential of massively disturbing existing game worlds. Something that always causes civil-war-like conditions in the community. Although I’d like to implement both I have a distinguished fear of pitchforks and torches
we’ll chase you with pitchforks, and cover you with tar and feathers anyway if you don’t come up with a solution. I don’t think anyone likes to nag you all the time, just like you probably don’t enjoy being nagged. But do you actually play the game ? I have checked and timed a few things on the Tempelhof server (writing this post has taken me at least half an hour)…
1 minute and 20 seconds to open my enterprise overview page,
4 minutes and 15 seconds to switch from the overview tab to the facts and figures tab,
4 minutes and 30 seconds to open the offices page.
It even took 15 seconds to open the game status page, to read that the server delay has now increased to 4 hours and 20 minutes. And by the way, we haven’t been able to use the ORS since friday.
I like this game, but waiting several minutes until a page shows up makes it unplayable.
Or am I the only guy who has to wait ages for some pages to open ?
The other two are clearly related to our statistics problems which are also the reason why we switched off the some of the global statistics for now. It’s a dilema, really: It’s one of the coolest features in AS but it’s causing a heck of trouble and we do not really know how to fix it because it’s just te nature of things that we need to deal with Gigs of data for this kind of statistics.
FWIW, I also wait ages for those same pages. RIght now (i.e., as of my check-in on the server a couple minutes ago) I’m just storing up available airplanes (sans flight plans) because the game is working too slowly for me to go through the wait times associated with assigning flights.
thanks for your reply. I would feel like an idiot if everybody else had no problems with long response times.
And indeed Martin, at this moment my overview page opens in a second. I don’t know why it took so long yesterday. But pages in general take too long to open. And I completely understand Outlier’s problem. Making or changing a flight schedule needs several mouse clicks and it is frustrating if every click means you have to wait half a minute or longer.
And by the way, Tempelhof is now 7 hours behind schedule.
Anyway, I am not an enemy of AS. I am an enthusiast and dedicated player. You guys have made an excellent and challenging game. But the game is drowning in the huge amount of data.
Getting rid of the ground network might cause trouble, but as far as I’m concerned: I prefer trouble instead of frustration. I won’t complain about reducing the ground network at all if it helps with the performance!
You mentioned scheduling flight was a tedious effort. How so? You get long response times in the actual process of adding flights? That’s weird, gotta look into this.
I’m lucky as I live in Australia so play during off peak, however one thing that takes a lot of time while adding flights, is the check buttons for days.
I think perhaps disconnecting the immediate processing of these may help solve server load slightly as you only process the request once, not 4 times.
On a side note, the next available slot button may also reduce human created demand by only making 1 call, not 20 (while searching for next available).
My loading times overall are very good, it is mainly just adding flights. At the risk of causing my self grief, perhaps rescheduling some of the regular processes to a time that is less used by human processes - quick temporary fix to give you guys some time to figure out underlying problems and solutions.
Speaking from a far out angle, ever considered running the servers SETI style? Players willing to leave their comps on to help process things do, while their machines are idle of course
I wouldn’t rule out the possibility that some of the cool new features (signature link, automatic route maps) create quite some load. For example, if signature links (which are generated on the fly, AFAIK) are used in high traffic websites you get a lot of database queries that are not caused by actually playing the game.
We have caching in place for that and the images are usually delivered within a few milliseconds. Also, we have an eye on the traffic from signature images and so far they do not cause any noticeable load on the servers.
Another question concerning scheduling performance: Does it affect you on all game worlds or just the "trouble makers" (like Tempelhof or Devau, although the latter seems to work ok recently)?
As mentioned on the dev-log, I intended to do something about scheduling-performance by tweaking the handling of slot-matrices, but unfortunately this didn’t pan out so far.
I am playing on Tempelhof and Stapleton, and both have regularly been affected during the last week. The current pattern on Stapleton is that the server has a backlog of up to 1 hour during most of the day (currently it’s 20 minutes). The ORS is rarely available. Response times get very long in the afternoon and stay that way until late night. The only time of the day where it’s fun to play is in the morning (after the backlog was processed and before the load increases again).
it is hard to tell. Sometimes speed is okay. Well, it’s always like talking on the phone with someone across the world. There is always a delay of a few seconds. And sometimes you edit a flight, change the time, wait, and wait… until the flight actually jumps to the new time.
The same happens when you switch from the scheduling page to the flight list. My gut feeling says that Stapleton is a bit faster. But gut feeling is not an objective way to measure
We will most likely move THF to dedicated hardware before the end of the year so we can be sure it’s not Stapleton and Tempelhof slowing each other down on the same box.
Because to be honest: Even though we know it’s bad design that’s causing most of the issues, I have no other solution at hand than throwing more expensive gear at the problem. There goes my summer vacation
