ZOS, tell the community whether or not a server hardware solution could fix Cyrodiil performance.

MurderMostFoul

To my knowledge, ZOS has never clearly stated that hardware solutions on their wouldn't improve performance in Cyrodiil. If they have, please share it here so myself and others can stop wondering about this.

ZOS,

I fully support efforts to reduce lag in Cyrodiil. I appreciate the time and the effort it takes to design solutions and test them. But before you radically change your game to help performance, please let the community if you can't achieve performance improvements with hardware upgrades. Countless comments have been posted, saying, "just get new servers." And for each of these comments, there are several more community members wondering if hardware upgrades could work. Tell us, and put this issue to rest. I, and likely many, many more, would rally in support of your efforts to fix Cyrodiil through sweeping combat changes, if we knew that server upgrades couldn't do the same thing. If hardware might help, but there are other limitations that make that option unfeasible, explain it to us. Leaving us in the dark regarding performance improvements through hardware upgrades just leads the community to speculate about your priorities. As long as this question is unanswered, the community will wonder if you are fixing the game by taking parts of it away, rather than investing in it, penalizing players for the sake of profits.

LightningWitch

It's definitely not a hardware issue. Amazon Web Services are used for the servers in ESO, and I can attest the issue cannot be hardware related.

AWS will scale based on need, meaning if an instance is at capacity, the server size will automatically grow to accommodate.

PVP is a different beast and I wouldn't be surprised if a completely different configuration is used due to the volume of players in PVP vs PVE.

Given this situation, it's only logical to deduce the issue is software related, incapable to updating fast enough to compensate for the action going on screen. Lag is the result, followed by disconnects.

I will set aside one common complaint: animations. This isn't the issue. Most of those are already in their own memory space, and since they're reused (which is why everything looks identical), calls to render are reduced.

If I were to make an educated guess at the issue, it's the skills. With so many having such high levels of calculations and timed effects, trying to calculate so many in a single instance is pushing the server to the point trying to render them instantly is difficult to do.

One thing I believe may be helpful, though it probably won't be well-liked: limit skills to a single bar.

MurderMostFoul

LightningWitch wrote: »

It's definitely not a hardware issue. Amazon Web Services are used for the servers in ESO, and I can attest the issue cannot be hardware related.

AWS will scale based on need, meaning if an instance is at capacity, the server size will automatically grow to accommodate.

PVP is a different beast and I wouldn't be surprised if a completely different configuration is used due to the volume of players in PVP vs PVE.

Given this situation, it's only logical to deduce the issue is software related, incapable to updating fast enough to compensate for the action going on screen. Lag is the result, followed by disconnects.

I will set aside one common complaint: animations. This isn't the issue. Most of those are already in their own memory space, and since they're reused (which is why everything looks identical), calls to render are reduced.

If I were to make an educated guess at the issue, it's the skills. With so many having such high levels of calculations and timed effects, trying to calculate so many in a single instance is pushing the server to the point trying to render them instantly is difficult to do.

One thing I believe may be helpful, though it probably won't be well-liked: limit skills to a single bar.

I appreciate your explanation, and acknowledge that many community members may have expertise in these matters, in contrast to my dearth of networking/server infrastructure knowledge. However, it should not be left to community members to explain to others their hypotheses regarding the viability of hardware solutions. It should come from ZOS, and end all speculation.

redshirt_49

Yeah, most common complaint I've heard is "Just upgrade the servers".

It really, REALLY is not that simple.

And even if it were, such an upgrade would cost a lot for the initial upgrade and for renting server space in the long run.
Gonna go ahead and say the suits aren't going to sign off on this.

Elsonso

LightningWitch wrote: »

It's definitely not a hardware issue. Amazon Web Services are used for the servers in ESO, and I can attest the issue cannot be hardware related.

No. I am pretty sure that the megaservers are on-prem, not cloud.

Davor

MurderMostFoul wrote: »

To my knowledge, ZOS has never clearly stated that hardware solutions on their wouldn't improve performance in Cyrodiil. If they have, please share it here so myself and others can stop wondering about this.

This will never happen. Zenimax (or Bethesda or what ever was before Bethesda) was made from layers. One thing layers know is don't admit to anything that can be used against you in court.

Also if Zenimax said it was hardware related, then Amazon can sue Zenimax for slander. Lawyers know the ins and outs and will never ever admit to something their fault so a class action lawsuit can be brought against them.

MurderMostFoul

redshirt_49 wrote: »

Yeah, most common complaint I've heard is "Just upgrade the servers".

It really, REALLY is not that simple.

And even if it were, such an upgrade would cost a lot for the initial upgrade and for renting server space in the long run.
Gonna go ahead and say the suits aren't going to sign off on this.

If it is truly prohibitively expensive, let's hear it from ZOS.

redshirt_49

MurderMostFoul wrote: »

redshirt_49 wrote: »

Yeah, most common complaint I've heard is "Just upgrade the servers".

It really, REALLY is not that simple.

And even if it were, such an upgrade would cost a lot for the initial upgrade and for renting server space in the long run.
Gonna go ahead and say the suits aren't going to sign off on this.

If it is truly prohibitively expensive, let's hear it from ZOS.

You don't honestly believe that the dev team is allowed to speak ill of their bosses in public and be allowed to keep their jobs? Hah, yeah that's not going to happen.

MurderMostFoul

Davor wrote: »

MurderMostFoul wrote: »

To my knowledge, ZOS has never clearly stated that hardware solutions on their wouldn't improve performance in Cyrodiil. If they have, please share it here so myself and others can stop wondering about this.

This will never happen. Zenimax (or Bethesda or what ever was before Bethesda) was made from layers. One thing layers know is don't admit to anything that can be used against you in court.

Also if Zenimax said it was hardware related, then Amazon can sue Zenimax for slander. Lawyers know the ins and outs and will never ever admit to something their fault so a class action lawsuit can be brought against them.

While I'm not a software/network engineer, I happen to be an attorney.

If ZOS is unwilling to comment on this, it is not due to potential liability.

Elsonso

MurderMostFoul wrote: »

redshirt_49 wrote: »

Yeah, most common complaint I've heard is "Just upgrade the servers".

It really, REALLY is not that simple.

And even if it were, such an upgrade would cost a lot for the initial upgrade and for renting server space in the long run.
Gonna go ahead and say the suits aren't going to sign off on this.

If it is truly prohibitively expensive, let's hear it from ZOS.

Actually, my assumption is that it is actually less expensive to "solve this with hardware" and that this is part of what they have been doing for the last couple of years. (Edit: meaning they are at the hardware cap. There is no source of more powerful hardware that they can use, at least not without tossing out everything and moving off of commercial server platforms)

MurderMostFoul

redshirt_49 wrote: »

MurderMostFoul wrote: »

redshirt_49 wrote: »

Yeah, most common complaint I've heard is "Just upgrade the servers".

It really, REALLY is not that simple.

And even if it were, such an upgrade would cost a lot for the initial upgrade and for renting server space in the long run.
Gonna go ahead and say the suits aren't going to sign off on this.

If it is truly prohibitively expensive, let's hear it from ZOS.

You don't honestly believe that the dev team is allowed to speak ill of their bosses in public and be allowed to keep their jobs? Hah, yeah that's not going to happen.

No one needs to speak ill of anyone if there is a reasonable explanation for not deploying a hardware solution. ZOS's silence on the matter leads us to wonder if there isn't a reasonable explanation.

Elsonso

redshirt_49 wrote: »

MurderMostFoul wrote: »

redshirt_49 wrote: »

Yeah, most common complaint I've heard is "Just upgrade the servers".

It really, REALLY is not that simple.

And even if it were, such an upgrade would cost a lot for the initial upgrade and for renting server space in the long run.
Gonna go ahead and say the suits aren't going to sign off on this.

If it is truly prohibitively expensive, let's hear it from ZOS.

You don't honestly believe that the dev team is allowed to speak ill of their bosses in public and be allowed to keep their jobs? Hah, yeah that's not going to happen.

Lambert can almost say what he wants. He's the Creative Director for the game, and that is not a grunt dev position. Previous Creative Directors at ZOS have talked about whether hardware could fix Cyrodiil performance, so Lambert could do it, if he wanted to. He does not want to. It is not information that he feels is an important amendment to his press release.

By the way, when Paul Sage answered this question, the answer was that it was not something that could be fixed by hardware.

Kadoin

It's not a hardware issue, it's definitely the checks and sheer amount of floating point in the game, but not all is going to be fixed by limiting checks and floating point in the way ZOS has indicated. I have mentioned it many times, but here was one response made just this year in another thread:

Here is an example of the damage drop bug that happens with speed I posted some details about it, but obviously not how to directly replicate it here:

Kadoin wrote: »

I can still bug all damage against me with 30 percent movement speed on live...

The problem is not a speed bug, but a physics bug that occurs when the player position value is truncated. I am guessing their collision system checks player position and uses exact value matching rather than a rounded value on floating point. That means if you know where to move in certain positions on the screen relative to players you can always avoid damage at any distance, movement speed just gives you time to move into that position better because it adds a static delay to skill transit time once you move a certain distance (another bug or intended?). This causes damage to drop because the skill thinks you are in one place, but your positional value is actually referencing another (because its off by a few decimal places, really ZOS?).

What they need to do is round the values in the positioning system to integer IF ONLY for that check. As long as they don't it don't matter what speed someone is moving if they know where to move relative to their opponent to trigger this bug, or the terrain locations in Cyro that also trigger this bug. This would also fix some bugs in PvE that would appear to be due to lag but aren't. However, it would mean that movement speed abusers now will come to the forums and rage...

But don't don't worry, I am sure ZOS knows what they are doing After all, there is no one around here that knows anyone that abuses these two bugs 100 percent of the time right?

...Right?

EDIT: I should also mention that that is all derived from randomly triggering the bug + testing to replicate it. And yes, no damage text or any "miss" or "dodge" occurs when that bug does.

Someone actually posted a video where their body was in another play but they were still able to kill others without being targeted recently on the forums that confirms this is the case... That kind of bug won't be fixed by "improving performance" but by rewriting the game's core systems.

Here's another bug that I reported and assume they have patched by now described, once again, it has to do with floating point. I actually predicted it would happen before even being able to test it, and it worked as expected:

Kadoin wrote: »

You would be able to do it if you had knowledge of computer science...though some bugs are very obvious. That same knowledge helps you to think of ways to break the game, and more often than not, you will be able to break the game with bugs that stem from common mistakes in computer programming.

Ex. Elfbane + BSW causing BSW's burn to infinistack is a bug I predicted because:

ZOS uses floating point values on skill duration and doesn't floor or ceil the values yet, meaning there is a case where a comparison can bug out if the floating-point value truncates and an exact value is required, and/or a larger value can enter the variable they use to track duration than the variable supports. That's literally the most likely reason this bug occurs, but hey I am not on the dev team so I have no idea what code issue exists that causes it, but I do know it is bugged and that was my method for guessing the bug would exist before it was even found, and it does exist and works as predicted.

Truncation of floating values actually seems to cause many bugs in this game believe it or not. I cannot say any more because I wouldn't want to help an abuser find out the more serious bugs that have not been fixed in years...(maybe not possible)

But someone else posted "why would you care?" He is right, unless you like that kind of thing, there is no reason to care because you gain nothing from reporting bugs and if you abuse them you will get banned...

And another example, ironically, a prediction that if light turned single target, spamming it would contribute to the massive lag because of increased calculations and checks:

Kadoin wrote: »

BohnT wrote: »

Kadoin wrote: »

BohnT wrote: »

The more people you face the worse it get and the more people you have fighting with you it gets so mich better.

I want this changed into this: -x^2/16+1

Not sure if serious or trolling. Worst balance suggestion I have ever seen on the forum. Have you even done the math or are you pulling that out of thin air? And I'm sure the servers need to compute more floating point when they are already slow. Guess we will have extreme lag when enough lights are placed, right?

Look at the graph of exp(x) and -x^2/16 +1 with the x and y axis i described in my post.
There's not much math involved in this also it has nothing to do with the server because that's just the skill performance vs different amounts of enemies and that can be done for every skill in this game.

Here's my problem: you want them to add a feature to a skill that a minority of players even whine about in a way that will only be useful if you're facing idiots doing bow light attacks. If it's added to backlash, there is no reason any other skill shouldn't receive similar scaling. If they do that, I guarantee it will result in more lag because I'm certain the server calculates damage and it will result in more floating point calculations. I know for sure I will start using my templar again just to cause lag by spamming light on entire zergs But yeah, dismiss that so you can pretend you are going to live longer under backlash, even though all the other skills will still kill you just as fast in Xv1 OR alternatively, the skill wont be used and will be replaced with a CC on their bar (better choice 99.999% of the time anyway). Even if not, they will move to the next "Xv1 tool" which is every damage skill in the game.

I'm not going to even lie, I have less problems with backlash deaths than any other skill in the game that has a CC or interrupt attached. I'm dead serious about that statement, and this is on ANY class I play. I routinely encounter zergs and groups of people all the time in Cyrodil and backlash is the LAST of my worries. The fact that you don't talk about interrupt spam, you don't talk about purifying, and talk about LoS means that you're likely a stam build...Let me take a magic guess, stam NB? If so, keep the jokes coming about what's balanced and what's not.

Keep in mind these posts were from different patches over the years, and yet many of them are still valid.

I am certain there are far more posts I have made on performance, but the search function is pretty hard to use on this forum for me I can't say I didn't see these changes coming, but I can say they aren't the "end-all" changes needed to increase performance.

Examples like these lead me to believe that ZOS won't really accomplish everything they are setting out to do with the change, and while it should and most likely improve performance to an extent, conditions where bugs such as the two described above will still cause the server to get bogged down and gain unnecessary stress, like conditions that happen when player position values start to truncate (this does happen to cause massive lag, and is triggered by playing normally sometimes. If your position is invalid and you roll dodge or "miss and dodge an attack" the server starts to choke and not only drop damage, but it appears to lag the more actions taken against you in that state).

The physics and positioning systems both use floating point, and ZOS will have to address that eventually.

x48rph

It's just not a problem that throwing more hardware at it can fix. At the end of the day calculations and all the back and forth communications can only happen so fast. That's why optimizations are so important. You can have 100 core processor but the other 99 are going to be idle if they all have to wait on one to finish first. But even then only so much can be done to optimize things. Eventually it will be too much and some processes will just be stuck waiting. The current crisis in cyrodil is just exposing the limits of their code and the underlying engine. While the solution might not be ideal, at least they are trying to fix it which is better then just continuing to ignore it.

vamp_emily

Kadoin wrote: »

It's not a hardware issue, it's definitely the checks and sheer amount of floating point in the game, but not all is going to be fixed by limiting checks and floating point in the way ZOS has indicated. I have mentioned it many times, but here was one response made just this year in another thread:

Here is an example of the damage drop bug that happens with speed I posted some details about it, but obviously not how to directly replicate it here:

Kadoin wrote: »

I can still bug all damage against me with 30 percent movement speed on live...

The problem is not a speed bug, but a physics bug that occurs when the player position value is truncated. I am guessing their collision system checks player position and uses exact value matching rather than a rounded value on floating point. That means if you know where to move in certain positions on the screen relative to players you can always avoid damage at any distance, movement speed just gives you time to move into that position better because it adds a static delay to skill transit time once you move a certain distance (another bug or intended?). This causes damage to drop because the skill thinks you are in one place, but your positional value is actually referencing another (because its off by a few decimal places, really ZOS?).

What they need to do is round the values in the positioning system to integer IF ONLY for that check. As long as they don't it don't matter what speed someone is moving if they know where to move relative to their opponent to trigger this bug, or the terrain locations in Cyro that also trigger this bug. This would also fix some bugs in PvE that would appear to be due to lag but aren't. However, it would mean that movement speed abusers now will come to the forums and rage...

But don't don't worry, I am sure ZOS knows what they are doing After all, there is no one around here that knows anyone that abuses these two bugs 100 percent of the time right?

...Right?

EDIT: I should also mention that that is all derived from randomly triggering the bug + testing to replicate it. And yes, no damage text or any "miss" or "dodge" occurs when that bug does.

Someone actually posted a video where their body was in another play but they were still able to kill others without being targeted recently on the forums that confirms this is the case... That kind of bug won't be fixed by "improving performance" but by rewriting the game's core systems.

Here's another bug that I reported and assume they have patched by now described, once again, it has to do with floating point. I actually predicted it would happen before even being able to test it, and it worked as expected:

Kadoin wrote: »

You would be able to do it if you had knowledge of computer science...though some bugs are very obvious. That same knowledge helps you to think of ways to break the game, and more often than not, you will be able to break the game with bugs that stem from common mistakes in computer programming.

Ex. Elfbane + BSW causing BSW's burn to infinistack is a bug I predicted because:

ZOS uses floating point values on skill duration and doesn't floor or ceil the values yet, meaning there is a case where a comparison can bug out if the floating-point value truncates and an exact value is required, and/or a larger value can enter the variable they use to track duration than the variable supports. That's literally the most likely reason this bug occurs, but hey I am not on the dev team so I have no idea what code issue exists that causes it, but I do know it is bugged and that was my method for guessing the bug would exist before it was even found, and it does exist and works as predicted.

Truncation of floating values actually seems to cause many bugs in this game believe it or not. I cannot say any more because I wouldn't want to help an abuser find out the more serious bugs that have not been fixed in years...(maybe not possible)

But someone else posted "why would you care?" He is right, unless you like that kind of thing, there is no reason to care because you gain nothing from reporting bugs and if you abuse them you will get banned...

And another example, ironically, a prediction that if light turned single target, spamming it would contribute to the massive lag because of increased calculations and checks:

Kadoin wrote: »

BohnT wrote: »

Kadoin wrote: »

BohnT wrote: »

The more people you face the worse it get and the more people you have fighting with you it gets so mich better.

I want this changed into this: -x^2/16+1

Not sure if serious or trolling. Worst balance suggestion I have ever seen on the forum. Have you even done the math or are you pulling that out of thin air? And I'm sure the servers need to compute more floating point when they are already slow. Guess we will have extreme lag when enough lights are placed, right?

Look at the graph of exp(x) and -x^2/16 +1 with the x and y axis i described in my post.
There's not much math involved in this also it has nothing to do with the server because that's just the skill performance vs different amounts of enemies and that can be done for every skill in this game.

Here's my problem: you want them to add a feature to a skill that a minority of players even whine about in a way that will only be useful if you're facing idiots doing bow light attacks. If it's added to backlash, there is no reason any other skill shouldn't receive similar scaling. If they do that, I guarantee it will result in more lag because I'm certain the server calculates damage and it will result in more floating point calculations. I know for sure I will start using my templar again just to cause lag by spamming light on entire zergs But yeah, dismiss that so you can pretend you are going to live longer under backlash, even though all the other skills will still kill you just as fast in Xv1 OR alternatively, the skill wont be used and will be replaced with a CC on their bar (better choice 99.999% of the time anyway). Even if not, they will move to the next "Xv1 tool" which is every damage skill in the game.

I'm not going to even lie, I have less problems with backlash deaths than any other skill in the game that has a CC or interrupt attached. I'm dead serious about that statement, and this is on ANY class I play. I routinely encounter zergs and groups of people all the time in Cyrodil and backlash is the LAST of my worries. The fact that you don't talk about interrupt spam, you don't talk about purifying, and talk about LoS means that you're likely a stam build...Let me take a magic guess, stam NB? If so, keep the jokes coming about what's balanced and what's not.

Keep in mind these posts were from different patches over the years, and yet many of them are still valid.

I am certain there are far more posts I have made on performance, but the search function is pretty hard to use on this forum for me I can't say I didn't see these changes coming, but I can say they aren't the "end-all" changes needed to increase performance.

Examples like these lead me to believe that ZOS won't really accomplish everything they are setting out to do with the change, and while it should and most likely improve performance to an extent, conditions where bugs such as the two described above will still cause the server to get bogged down and gain unnecessary stress, like conditions that happen when player position values start to truncate (this does happen to cause massive lag, and is triggered by playing normally sometimes. If your position is invalid and you roll dodge or "miss and dodge an attack" the server starts to choke and not only drop damage, but it appears to lag the more actions taken against you in that state).

The physics and positioning systems both use floating point, and ZOS will have to address that eventually.

You seem like a very smart person. +1

Four_Fingers

Meanwhile using TCP for transport rather than UDP is the white elephant in the room.

LightningWitch

MurderMostFoul wrote: »

I appreciate your explanation, and acknowledge that many community members may have expertise in these matters, in contrast to my dearth of networking/server infrastructure knowledge. However, it should not be left to community members to explain to others their hypotheses regarding the viability of hardware solutions. It should come from ZOS, and end all speculation.

Let's review your words for a second. You clearly admitted to having knowledge in this arena, yet you're demanding ZoS to confirm what you already know?

Not going to happen.

If the NDA isn't enough to stop employees from disclosing the information, then common sense will definitely play its part here because no employee needs to confirm anything to you.

I don't know how long you've been playing the game, but I can tell you there are two options open to you: deal with the issues, and hope they're resolved, or leave the game and find something more suitable to your expectations.

If you do decide to choose the latter and decide to come back, you will not find me surprised your reply will be "It's like this everywhere."

I've been gaming for many years, and to this day, I've never found a PvP game without lag or other issues. It doesn't matter if the server is dedicated, shared, or on site: pushing so much math in a little time is always going to equate to lag.

No matter how fast a processor is, time is always need to perform calculations.

On that, good luck with your new quest, but don't get upset when you find it bugged and you can't complete it.

EmEm_Oh

Davor wrote: »

MurderMostFoul wrote: »

To my knowledge, ZOS has never clearly stated that hardware solutions on their wouldn't improve performance in Cyrodiil. If they have, please share it here so myself and others can stop wondering about this.

This will never happen. Zenimax (or Bethesda or what ever was before Bethesda) was made from layers. One thing layers know is don't admit to anything that can be used against you in court.

Also if Zenimax said it was hardware related, then Amazon can sue Zenimax for slander. Lawyers know the ins and outs and will never ever admit to something their fault so a class action lawsuit can be brought against them.

And then ZOS could provide evidence, proving it was so...then what.

There's nothing wrong with being honest about what's happening. That would not be slander or libel. It would be a proven fact. And if ZOS said it was X issue or Y issue, I would hope they would have those facts to back up their claims anyway.

But yeah, it's becoming more clear it's not hardware per se. It's got everything, or mostly everything, to do with software calculations per second. This may not be technically a CPU issue, because if you have out-dated software code to begin with, the software makes calculations that put more strain than necessary on the final result if it has to toss out the code that is either old or just "doesn't make sense". This is why people end up below the ground, or when Shooting Star hits you on a wall, you can get pushed into it from the explosion, etc.

ZOS must find a way to allow all these calculations or quite frankly, ESO is done. I'm not sure what other MMOs are doing, but if ESO is to expand into a world that is so amazingly large, extremely exciting whenever one logs on due to the massive amount of possibilities, and virtually endless exploration...they MUST solve this calculation problem. Because it only gets more complicated. More skills will be added...more classes...

This is a die or thrive point in ESO history. Nerfing AOEs is just another band-aid, imo. What else are they going to nerf in order to spare the calculations?

Dagoth_Rac

ZOS: "We could solve performance problems by adding more hardware resources."

Forums: "Then do it! Stop being cheapskates!"


ZOS: "We cannot solve performance problems by adding more hardware resources."

Forums: "Why should we believe you? You are probably just too cheap to add more hardware."

ShawnLaRock

It can’t.

S.

Edit: we’re placing too much emphasis on this one area, when the entire existing infrastructure can’t handle the intended calculations - and therefore, “balancing” is to those PHYSICAL limitations, rather than the interrelated math.

Simply - my internet is fine; their internet is often fine - the base game doesn’t know how to figure out how my set works with a bunch of other people’s, is all... and it’s a big deal. THE deal.

Kitrani

But... all companies just improve technical equipment instead of breaking game mechanics.

Casterial

Hardware and network code.
1) Hardware, this has been proven true during MYM. During MYM we have almost no lag, almost as if they up the server needs and hardware needed for ESO.
2) Network code, the game was never made to have these big numbers, the game was not made to incorporate CP calculations, the game was not made to calculate battle spirit....But yet it does now and numbers are x9.8 what they originally were.

Elsonso

Dagoth_Rac wrote: »

ZOS: "We could solve performance problems by adding more hardware resources."

Forums: "Then do it! Stop being cheapskates!"


ZOS: "We cannot solve performance problems by adding more hardware resources."

Forums: "Why should we believe you? You are probably just too cheap to add more hardware."

Forum: "You downgraded your servers, and now you want to completely redesign and nerf combat to avoid upgrading the servers to where they used to be. We demand ANSWERS!"

ZOS:

gatekeeper13

As we saw in the MYM, a hardware upgrade would definitely help.

daemonios

LightningWitch wrote: »

It's definitely not a hardware issue. Amazon Web Services are used for the servers in ESO, and I can attest the issue cannot be hardware related.

AWS will scale based on need, meaning if an instance is at capacity, the server size will automatically grow to accommodate.

PVP is a different beast and I wouldn't be surprised if a completely different configuration is used due to the volume of players in PVP vs PVE.

Given this situation, it's only logical to deduce the issue is software related, incapable to updating fast enough to compensate for the action going on screen. Lag is the result, followed by disconnects.

I will set aside one common complaint: animations. This isn't the issue. Most of those are already in their own memory space, and since they're reused (which is why everything looks identical), calls to render are reduced.

If I were to make an educated guess at the issue, it's the skills. With so many having such high levels of calculations and timed effects, trying to calculate so many in a single instance is pushing the server to the point trying to render them instantly is difficult to do.

One thing I believe may be helpful, though it probably won't be well-liked: limit skills to a single bar.

I'm not a hardware/networking expert, but I'd eat my hat (if I wore one) if you were proved right.

That the issues are at least partly hardware related, or can be mitigated, is demonstrated by certain events during which performance increases very markedly, then degrades as the event ends. You may be right about AWS and their ability to scale. It doesn't mean ZOS pays for unlimited scalability. More likely they contact a certain service level, which they vary when they deem it cost-effective. Permanently increasing the service level is probably not seen as being cost-effective.

PrimusNephilim

Davor wrote: »

This will never happen. Zenimax (or Bethesda or what ever was before Bethesda) was made from layers. One thing layers know is don't admit to anything that can be used against you in court.

layers...kinda like an onion, their servers are onions and that's why we have these issues....got it!
Get another onion please....there, problem solved!

MurderMostFoul

LightningWitch wrote: »

Let's review your words for a second. You clearly admitted to having knowledge in this arena, yet you're demanding ZoS to confirm what you already know?

Not going to happen.

I said I lack knowledge in this area.

LightningWitch wrote: »

If the NDA isn't enough to stop employees from disclosing the information, then common sense will definitely play its part here because no employee needs to confirm anything to you.

I'm not asking any employee to break an NDA (an agreement between individual employees and the company), I'm asking for transparency from the company. I know they don't need to, but if they have a reasonable explanation for why a hardware solution won't work, I believe it is in their best interest to share it.

LightningWitch wrote: »

I don't know how long you've been playing the game, but I can tell you there are two options open to you: deal with the issues, and hope they're resolved, or leave the game and find something more suitable to your expectations.

If you do decide to choose the latter and decide to come back, you will not find me surprised your reply will be "It's like this everywhere."

I've been gaming for many years, and to this day, I've never found a PvP game without lag or other issues. It doesn't matter if the server is dedicated, shared, or on site: pushing so much math in a little time is always going to equate to lag.

No matter how fast a processor is, time is always need to perform calculations.

I've been online gaming for close to 20 years. I know all games have their issues. But gaming companies, and more broadly all customer service companies, fair better with their customers by being transparent about their issues. Conversely, they may oberate opaquely when the truth behind the issues would be damaging if revealed.

zaria

LightningWitch wrote: »

It's definitely not a hardware issue. Amazon Web Services are used for the servers in ESO, and I can attest the issue cannot be hardware related.

AWS will scale based on need, meaning if an instance is at capacity, the server size will automatically grow to accommodate.

PVP is a different beast and I wouldn't be surprised if a completely different configuration is used due to the volume of players in PVP vs PVE.

Given this situation, it's only logical to deduce the issue is software related, incapable to updating fast enough to compensate for the action going on screen. Lag is the result, followed by disconnects.

I will set aside one common complaint: animations. This isn't the issue. Most of those are already in their own memory space, and since they're reused (which is why everything looks identical), calls to render are reduced.

If I were to make an educated guess at the issue, it's the skills. With so many having such high levels of calculations and timed effects, trying to calculate so many in a single instance is pushing the server to the point trying to render them instantly is difficult to do.

One thing I believe may be helpful, though it probably won't be well-liked: limit skills to a single bar.

An point but running on an huge server farm will not help you on an pop locked cyrodil campaign.
Simply as an instance has to run on one single server.
This is not much of an problem on overland instances. Here you want to keep player count down so you don't flood delves and quest locations anyway. Trials, bg and dungeon instances can be fit on spare capacity on any server.
The problem is the Cyrodil main campaigns or 6-9 servers, could they run them on much stronger servers than the standard ones, it depend on Amazon policy here. You can buy 8 cpu servers with over 200 cores but this will set you back over $100K
An 4 cpu with 100 cores for an 3rd of that. Even with the old 3x600 players each core only had to handle 16 players worst case for the 4 cpu version.
Now the ESO server system might not be fully multi threaded so it might not scale well, this will be more of an problem in large fights but you could insulate the players outside this fights better.

Who brings up the elephant in the room, the clients. In short if you get ESO problems in an Alkir dolmen runs or simply showing up at overpopulated locations like Brendas stand in midwinter festival pop locked PvP campaigns don't work for you.

No its not just people on craptops, some with good gaming setups has problems so its likely an software element too.
And why does Cyrodil has the same graphic settings as overland? Most configure their graphic for best quality then overland out of combat.

In short simply upgrading the 6-9 servers would probably be cheaper than the incoming mess. However its likely internal issues and might well be limits on the amazon system. Or it could be server software issues.
And the client issue might be something the ones making the client want to cover up, again if you can not do prime time zergs on Alkir dolmens pretty smoothly you have no place in an zerg on an pop locked campaign.
And players agree, its not their system faults.

Tommy_The_Gun

I am actually surprised that in their official statement, they did not stated that...
https://forums.elderscrollsonline.com/en/discussion/539136/update-on-cyrodiil-performance-upcoming-aoe-tests/p1

It is obvious they should. They could say like "anything". Like for example that they already use best hardware available money can buy etc. But they did not, and people think that they simply... don't want to invest money...

zaria

daemonios wrote: »

LightningWitch wrote: »

It's definitely not a hardware issue. Amazon Web Services are used for the servers in ESO, and I can attest the issue cannot be hardware related.

AWS will scale based on need, meaning if an instance is at capacity, the server size will automatically grow to accommodate.

PVP is a different beast and I wouldn't be surprised if a completely different configuration is used due to the volume of players in PVP vs PVE.

Given this situation, it's only logical to deduce the issue is software related, incapable to updating fast enough to compensate for the action going on screen. Lag is the result, followed by disconnects.

I will set aside one common complaint: animations. This isn't the issue. Most of those are already in their own memory space, and since they're reused (which is why everything looks identical), calls to render are reduced.

If I were to make an educated guess at the issue, it's the skills. With so many having such high levels of calculations and timed effects, trying to calculate so many in a single instance is pushing the server to the point trying to render them instantly is difficult to do.

One thing I believe may be helpful, though it probably won't be well-liked: limit skills to a single bar.

I'm not a hardware/networking expert, but I'd eat my hat (if I wore one) if you were proved right.

That the issues are at least partly hardware related, or can be mitigated, is demonstrated by certain events during which performance increases very markedly, then degrades as the event ends. You may be right about AWS and their ability to scale. It doesn't mean ZOS pays for unlimited scalability. More likely they contact a certain service level, which they vary when they deem it cost-effective. Permanently increasing the service level is probably not seen as being cost-effective.

I assume the performance increase during PvP events was because the large ball group guilds spread out over servers to farm emperors or core servers got flooded by casuals doing PvDoor.

No they did not rent higher capacity pvp server during the event. Now they probably made some general servers PvP overflow servers, and we know they rented more general servers for PC-EU as it got more players but that was not flipping an switch.
Se my above post about why more servers does not help an Cyrodil campaign but helps for high population other places.

Simple solution reduce AP gain on server population in an way more radical way than now.
It keeps the AP farmers off the servers for people who enjoy large scale PvP. The good ones will still earn plenty and you can have mechanisms to reward them.
But get people to go other places to zerg or do PvDoor.
But it will not solve the performance issues with ball groups crashing.

Four_Fingers

Does someone have a source actually proving who owns the data centers in Texas and Germany?
I see the Amazon claim but no source.