Server Performance on Invasion Day (see other thread)

This site uses cookies. By continuing to browse this site, you are agreeing to our Cookie Policy.

  • Lexinator wrote:

    jwhite179 wrote:

    Korn wrote:

    jwhite179 wrote:

    Korn wrote:

    Personally: no as I am not a tech guy myself.
    I'll try my best though:
    The underlying issue - in "layman" terms - is that Albion has fast paced combat with lots of different abilities. These generate a lot of information. If you have 10 players in a fight (such as in MOBA) information generated by 1 player needs to be transmitted to 9 others. That true for each player, in both directions, so unless my math is wrong that would be 2*10*9 = 180 information pathways. If you double the number of players, this more than doubles, to 2*20*19 = 760 (which is an increase by more than a factor of 4). In our case, we up to hundreds of players per zone - the information load created here is huge. Secondly, while you can share some of that load across different hardware devices (parallel processing) this cannot be done universally as the different CPUs that you share the load on will still need to come to a common understanding of what is happening, i.e. they will need to talk between each other. What we can do is get the most powerful CPUs on the market, optimize as much as we can (and there is certainly still room for improvement) and then determine the limit of the number of players that can be in a fight concurrently. Such a limit exits, and the main problem right now if that zones clearly allow more players to fight at the same time that what our systems currently can handle. This is made worse right now by the general server load (here, there is also room for optimization, i.e. the general game world population should have less impact on the particular fights inside certain zones and vice versa).

    To sum it up: there are clear pathways for us to take to further optimize. There is however a hard limit to how large fights can get in with an action-based combat system such as the one we use. Hence, there are also a game design challenge for us: we need to make sure that incentives for mega-blobs are reduced. We'll need to look at the zone queue / cap mechanic again. Intentionally trying to cap out a zone should not be a thing. As stated above, we'll provide more details early next week, both on the tech side of things (by our coders who are more qualified to talk about this than I am) and on how we are going to tackle these issues.
    So, while I'm not doubting this is true in whole- this logic makes me a little curious. In World of Warcraft, AB had 100v100. In GW2 there can be as much as 100v100. GW2 also has siege weapons that have AoE's that can cover an entire screen. Sight is not limited to a top down perspective.These do not cause a significant impact on gameplay in those MMO's. The abilities, AoE's, and general effects are far more detailed in those games as well. So how is it, that both of those games who have been out for 7+ years and able to pull it off from their start? Technology gets better every day, and if they were able to pull it off over seven years ago- then what's keeping SBI from maintaining that standard of gaming?
    We go far beyond 100 vs 100. Based on the formula above, 2*100*99 = 19,800. Our zone cap right now is 350, giving 2*350*349 = 244,300. That's more than 12-times as information intensive as a 100 vs 100 fight. For those of you more technically minded, our coders will share a more detailed statement about the technical situation with you next week.
    You're telling me, that the packet of information being sent in this game from one individual, to 349 other players- is larger, than a 100v100 of this?
    A cap of 200 vs a cap of 350. That seems to be the issue.
    Oh, my apologies. GW2 WvW is actually capped at 666 players for weekly reset, and drops to a 300-500 capacity for one map depending on server size. So they are nearly running double of what a 350 body ZvZ is capable of.
  • jwhite179 wrote:

    Lexinator wrote:

    jwhite179 wrote:

    Korn wrote:

    jwhite179 wrote:

    Korn wrote:

    Personally: no as I am not a tech guy myself.
    I'll try my best though:
    The underlying issue - in "layman" terms - is that Albion has fast paced combat with lots of different abilities. These generate a lot of information. If you have 10 players in a fight (such as in MOBA) information generated by 1 player needs to be transmitted to 9 others. That true for each player, in both directions, so unless my math is wrong that would be 2*10*9 = 180 information pathways. If you double the number of players, this more than doubles, to 2*20*19 = 760 (which is an increase by more than a factor of 4). In our case, we up to hundreds of players per zone - the information load created here is huge. Secondly, while you can share some of that load across different hardware devices (parallel processing) this cannot be done universally as the different CPUs that you share the load on will still need to come to a common understanding of what is happening, i.e. they will need to talk between each other. What we can do is get the most powerful CPUs on the market, optimize as much as we can (and there is certainly still room for improvement) and then determine the limit of the number of players that can be in a fight concurrently. Such a limit exits, and the main problem right now if that zones clearly allow more players to fight at the same time that what our systems currently can handle. This is made worse right now by the general server load (here, there is also room for optimization, i.e. the general game world population should have less impact on the particular fights inside certain zones and vice versa).

    To sum it up: there are clear pathways for us to take to further optimize. There is however a hard limit to how large fights can get in with an action-based combat system such as the one we use. Hence, there are also a game design challenge for us: we need to make sure that incentives for mega-blobs are reduced. We'll need to look at the zone queue / cap mechanic again. Intentionally trying to cap out a zone should not be a thing. As stated above, we'll provide more details early next week, both on the tech side of things (by our coders who are more qualified to talk about this than I am) and on how we are going to tackle these issues.
    So, while I'm not doubting this is true in whole- this logic makes me a little curious. In World of Warcraft, AB had 100v100. In GW2 there can be as much as 100v100. GW2 also has siege weapons that have AoE's that can cover an entire screen. Sight is not limited to a top down perspective.These do not cause a significant impact on gameplay in those MMO's. The abilities, AoE's, and general effects are far more detailed in those games as well. So how is it, that both of those games who have been out for 7+ years and able to pull it off from their start? Technology gets better every day, and if they were able to pull it off over seven years ago- then what's keeping SBI from maintaining that standard of gaming?
    We go far beyond 100 vs 100. Based on the formula above, 2*100*99 = 19,800. Our zone cap right now is 350, giving 2*350*349 = 244,300. That's more than 12-times as information intensive as a 100 vs 100 fight. For those of you more technically minded, our coders will share a more detailed statement about the technical situation with you next week.
    You're telling me, that the packet of information being sent in this game from one individual, to 349 other players- is larger, than a 100v100 of this?
    A cap of 200 vs a cap of 350. That seems to be the issue.
    Oh, my apologies. GW2 WvW is actually capped at 666 players for weekly reset, and drops to a 300-500 capacity for one map depending on server size. So they are nearly running double of what a 350 body ZvZ is capable of.
    On the same engine?
  • jwhite179 wrote:

    Korn wrote:

    jwhite179 wrote:

    Korn wrote:

    Personally: no as I am not a tech guy myself.
    I'll try my best though:
    The underlying issue - in "layman" terms - is that Albion has fast paced combat with lots of different abilities. These generate a lot of information. If you have 10 players in a fight (such as in MOBA) information generated by 1 player needs to be transmitted to 9 others. That true for each player, in both directions, so unless my math is wrong that would be 2*10*9 = 180 information pathways. If you double the number of players, this more than doubles, to 2*20*19 = 760 (which is an increase by more than a factor of 4). In our case, we up to hundreds of players per zone - the information load created here is huge. Secondly, while you can share some of that load across different hardware devices (parallel processing) this cannot be done universally as the different CPUs that you share the load on will still need to come to a common understanding of what is happening, i.e. they will need to talk between each other. What we can do is get the most powerful CPUs on the market, optimize as much as we can (and there is certainly still room for improvement) and then determine the limit of the number of players that can be in a fight concurrently. Such a limit exits, and the main problem right now if that zones clearly allow more players to fight at the same time that what our systems currently can handle. This is made worse right now by the general server load (here, there is also room for optimization, i.e. the general game world population should have less impact on the particular fights inside certain zones and vice versa).

    To sum it up: there are clear pathways for us to take to further optimize. There is however a hard limit to how large fights can get in with an action-based combat system such as the one we use. Hence, there are also a game design challenge for us: we need to make sure that incentives for mega-blobs are reduced. We'll need to look at the zone queue / cap mechanic again. Intentionally trying to cap out a zone should not be a thing. As stated above, we'll provide more details early next week, both on the tech side of things (by our coders who are more qualified to talk about this than I am) and on how we are going to tackle these issues.
    So, while I'm not doubting this is true in whole- this logic makes me a little curious. In World of Warcraft, AB had 100v100. In GW2 there can be as much as 100v100. GW2 also has siege weapons that have AoE's that can cover an entire screen. Sight is not limited to a top down perspective.These do not cause a significant impact on gameplay in those MMO's. The abilities, AoE's, and general effects are far more detailed in those games as well. So how is it, that both of those games who have been out for 7+ years and able to pull it off from their start? Technology gets better every day, and if they were able to pull it off over seven years ago- then what's keeping SBI from maintaining that standard of gaming?
    We go far beyond 100 vs 100. Based on the formula above, 2*100*99 = 19,800. Our zone cap right now is 350, giving 2*350*349 = 244,300. That's more than 12-times as information intensive as a 100 vs 100 fight. For those of you more technically minded, our coders will share a more detailed statement about the technical situation with you next week.
    You're telling me, that the packet of information being sent in this game from one individual, to 349 other players- is larger, than a 100v100 of this?
    Not even just WoW, but there are other 3D games, some with much more intensive information transfer, such as first person shooters, that still manage to pull off dozens, if not hundreds, of players fighting each other seamlessly and flawlessly. The amount of information in this game, in a fixed-position isometric game.. It just cannot be even close to the same amount of information as a 3D battle featuring the same numbers.
  • jwhite179 wrote:

    Korn wrote:

    jwhite179 wrote:

    Korn wrote:

    Personally: no as I am not a tech guy myself.
    I'll try my best though:
    The underlying issue - in "layman" terms - is that Albion has fast paced combat with lots of different abilities. These generate a lot of information. If you have 10 players in a fight (such as in MOBA) information generated by 1 player needs to be transmitted to 9 others. That true for each player, in both directions, so unless my math is wrong that would be 2*10*9 = 180 information pathways. If you double the number of players, this more than doubles, to 2*20*19 = 760 (which is an increase by more than a factor of 4). In our case, we up to hundreds of players per zone - the information load created here is huge. Secondly, while you can share some of that load across different hardware devices (parallel processing) this cannot be done universally as the different CPUs that you share the load on will still need to come to a common understanding of what is happening, i.e. they will need to talk between each other. What we can do is get the most powerful CPUs on the market, optimize as much as we can (and there is certainly still room for improvement) and then determine the limit of the number of players that can be in a fight concurrently. Such a limit exits, and the main problem right now if that zones clearly allow more players to fight at the same time that what our systems currently can handle. This is made worse right now by the general server load (here, there is also room for optimization, i.e. the general game world population should have less impact on the particular fights inside certain zones and vice versa).

    To sum it up: there are clear pathways for us to take to further optimize. There is however a hard limit to how large fights can get in with an action-based combat system such as the one we use. Hence, there are also a game design challenge for us: we need to make sure that incentives for mega-blobs are reduced. We'll need to look at the zone queue / cap mechanic again. Intentionally trying to cap out a zone should not be a thing. As stated above, we'll provide more details early next week, both on the tech side of things (by our coders who are more qualified to talk about this than I am) and on how we are going to tackle these issues.
    So, while I'm not doubting this is true in whole- this logic makes me a little curious. In World of Warcraft, AB had 100v100. In GW2 there can be as much as 100v100. GW2 also has siege weapons that have AoE's that can cover an entire screen. Sight is not limited to a top down perspective.These do not cause a significant impact on gameplay in those MMO's. The abilities, AoE's, and general effects are far more detailed in those games as well. So how is it, that both of those games who have been out for 7+ years and able to pull it off from their start? Technology gets better every day, and if they were able to pull it off over seven years ago- then what's keeping SBI from maintaining that standard of gaming?
    We go far beyond 100 vs 100. Based on the formula above, 2*100*99 = 19,800. Our zone cap right now is 350, giving 2*350*349 = 244,300. That's more than 12-times as information intensive as a 100 vs 100 fight. For those of you more technically minded, our coders will share a more detailed statement about the technical situation with you next week.
    You're telling me, that the packet of information being sent in this game from one individual, to 349 other players- is larger, than a 100v100 of this?
    Bro, techwise:
    What you see in a 100v100 or in a 200v200 is basically up to your computer. And it will most likely handle it pretty well.
    The problem is in the information sent every x miliseconds between each person (client). Which in Albion's case is more than this (as you said 100v100 of this instead of 350 people in Albion). These little packets of information that have to be distributed very quickly and frequently between a ton of people easily clog up the servers.
  • jwhite179 wrote:

    Oh, my apologies. GW2 WvW is actually capped at 666 players for weekly reset, and drops to a 300-500 capacity for one map depending on server size. So they are nearly running double of what a 350 body ZvZ is capable of.

    That's interesting. Do you have a link / further information on this so we can look into it?

    The only thing I could find was this video where the WvW cap is said to be around 70 players per side. If anybody has further information on this, it would be awesome if you could share it.
  • MEATCUP wrote:

    On the same engine?
    No, they use Havok and Ragdoll. Havok controls the destructible content like walls, gates and resources. Ragdoll for the environment and mobs. I believe that arena net uses their own engine based off the first game. I'm not arguing the validity of game engines being used though. My challenge was against the concept of data transfer between 1 player, XXX-players, and that other development companies have been doing it for quite some time. Most arguably the size of the data that is being transferred between players.
  • Something that i think that might help what about the timer on mob corps to be lowered so that if its not looted or once its been looted that it stays for 30 seconds rather then having a huge pile of corps inside an area.
    Also something to test and look into is the lag while running into pvp zone. When a group starts to flag up from running into this it lags the whole zone.
  • @Korn

    I appreciate the feedback you can give as always.

    I am wondering if this server bottleneck / throttling is caused by mass particle effects? I just know from games ive played and seen created in studio that particle effects of this magnitude can create unnecessary packets therefore overloading the servers and delaying response times.
  • I believe though that other games tackle the issue by creating instances of the same zone to disperse the player base. Its similar to how the cities in AO are currently handled where when players reach a certain threshold it just stops showing most of the other players.

    I don't think players want to see a game though where after x players enter a zone a new instance is created just for the new group. Splitting up the populations on each map however will be needed some how.

    @Korn
    One solution could be implementing Friendly fire between alliances as discussed before. Follow it up with a cap of 100 people from the same guild per map. (so in other words a 300 man guild could only zone up to 100 members into a cluster at one time).

    The two changes would force alliances to spread out there forces as well as reduce the effectiveness of just funneling alliance members into one massive 300 man guild.

    The post was edited 1 time, last by DoomRawrus ().

  • Korn wrote:

    jwhite179 wrote:

    Oh, my apologies. GW2 WvW is actually capped at 666 players for weekly reset, and drops to a 300-500 capacity for one map depending on server size. So they are nearly running double of what a 350 body ZvZ is capable of.
    That's interesting. Do you have a link / further information on this so we can look into it?

    The only thing I could find was this video where the WvW cap is said to be around 70 players per side. If anybody has further information on this, it would be awesome if you could share it.

    This is a clip from one of the fights. The reason, it's called WvWvW is that you have 3 servers fighting each other individually. You can see that the person who made this video has 42/50 in his party alone. The green dots on his minimap are people from his server that are not in his party. The server capacity is limited to another server (to an extent). If Server A brings 100 people, but server B only has 70 people to counter, then server B gets an outnumbered buff. Increased healing, damage, and reduced damage. In this fight alone you can see 300+ players. Evidently, If there are only 30 people from server A, then they place a queue to prevent Server B and Server C from bringing 100+ to counter server A.


    Regardless. this is 300+ fight in one small area. There are multiple castles, warcamps, and supply depots throughout this map. With fights spread out throughout the entire map, in one instance.

    The post was edited 2 times, last by jwhite179 ().

  • jwhite179 wrote:

    Korn wrote:

    jwhite179 wrote:

    Oh, my apologies. GW2 WvW is actually capped at 666 players for weekly reset, and drops to a 300-500 capacity for one map depending on server size. So they are nearly running double of what a 350 body ZvZ is capable of.
    That's interesting. Do you have a link / further information on this so we can look into it?
    The only thing I could find was this video where the WvW cap is said to be around 70 players per side. If anybody has further information on this, it would be awesome if you could share it.
    This is a clip from one of the fights. The reason, it's called WvWvW is that you have 3 servers fighting each other individually. You can see that the person who made this video has 42/50 in his party alone. The green dots on his minimap are people from his server that are not in his party. The server capacity is limited to another server (to an extent). If Server A brings 100 people, but server B only has 70 people to counter, then server B gets an outnumbered buff. Increased healing, damage, and reduced damage. In this fight alone you can see 300+ players. Evidently, If there are only 30 people from server A, then they place a queue to prevent Server B and Server C from bringing 100+ to counter server A.


    Regardless. this is 300+ fight in one small area. There are multiple castles, warcamps, and supply depots throughout this map. With fights spread out throughout the entire map, in one instance.
    Thanks for sharing. I just went over the video.

    To get a rough idea, I counted the allied players at 4:00, I get around 50 players here counting blue dots and green dots - assuming blue dots are the players in his party. There is a map zoom out at around 9:24, it also shows green dots and blue dots and it looks like the main fight was taking place at the castle in the center in which the guy who posted the video took part in. All in all, this still looks consistent with around 70-100 players per side as mentioned in the video above.
  • Again, there are 3 servers. So by saying it could be 100 per side equals out to at least 300. Hell, buy the game and hop in one of the top 6 servers, and X up. Their reset was last night, but should still be pretty poppin today. Yet again, the packet size of information that they are transmitting from 1 player to XXX-amount of players, is incredibly likely to be larger than a packet of data being transferred in Albion that has minimal graphics, small visual effects, and quite frankly a simple UI. We pretty much get a Red or Green indicator on the ground. every ability in GW2 is highly emphasized and visually detailed.
  • Maybe make mega alliances, and 50 players cap on map? Sandbox and everyhere cap... not good. Now must change WC system against 180 steps. I wish good ideas, because as for now is going into the disadvantage.

    When you cannot give content for large groups, create something in the end PROFITABLE(world boss - funny loot, HG 2v2, 5v5- VERY funny loot, solo of the box - funny loot), only what is now profitable it is RD...

    The post was edited 2 times, last by Shamtu ().

  • Waiting for official response from ur tech team. I'd like to hear what could be changed to improve current situation. Would f.e hiding characters / reducing spells effects globally, etc. help? Hidden characters + disabled spell effects = less information to send = smaller choke?

    If it could help even a little, u can make a player limit for a zone (200~) and if there are more people in a single zone, hide people's characters or skill effects.
  • Server issues aside, I think you should implement a "ZvZ setting" Option in the graphic settings, Because currently "switching" to zvz settings takes quite a while because you need to change a few options.
    This option could also get activated automatically if too many players are in a single zone. ( I know this wouldnt help server issues, but it would be a nice QOL change)