So I did some looking, a lot of post on servers were on the champion forums (DAMN YOU!!!!! I'll get to that pledge soon enough)) and then I read on this one forum which got closed cause there were other forums that touched based on it but I wasn't really able to find any forums that talked about what I actually wanted other than that, and it was closed so....
Anyways, from what I gathered off of the closed forum, it seems like we'll have 1 of each server type (RP/PvP/PvE), and then possibly one implication server I think he said (idk, I'd have to look back but I'm too lazy for that). So what does VR have to say on that? Any confirmation or changes? And based off of how many people first play, suppose it's a lot more than you guys expect and people can't play due to the servers being full, how quickly would you be able to get more servers up and running, and how would you get the already settled in player bases of the already established servers to migrate over to the new servers. Even if it only takes you guys a day or two, a lot can happen (escpecially seeing you guys are a group centered game), bonds can be quickly made, guilds made, and the likes. Splitting that up might not be the easiest thing to do and the growth of the new servers might take awhile, which will make it all the less more desirable. How would you guys handle that situation?
Furthermore, how many can be online at a given time on one server? Idk if you'd have the answers to that quite yet, that might be something to ask when you guys have the servers pretty much set up, but just curiosity do you have an estimate? With SWTOR, I'd say that theres usually +500ish (1000's pushing it, probably a lot less) on at a single time throughout the day on one of the larger servers, such as Harb and EH (Most of the other servers are pretty pitiful though). But those servers, as well as all the other ones, are usually on the 'Light' status, and maybe occasionally the 'Standard' during their more active times, so based off of that I'd think that they'd be maxed out around 3000 perhaps? Would you guys be around that, is that your goal? Or would you perhaps be looking for less space?
This is mostly just to VR thinking (Hell, it's been what? 2-3 years at least now? I'm sure this has crossed your minds and has been discussed tbh) but I also need to pass my time and I found myself wondering what VR has planned on this type of thing.
We have no idea Jacasta, it is impossible to even guess at this point in time, that is what testing will be used for, to gather the data and see what pop-to-server performance ratio is best and then we will go from there.
How many servers will depend on how many people play and what number we deem to be suitable for each server, we want them to feel lived in and social without being over populated and have queues for everything.
So this can't be answered until we get well into testing and have everything we need implemented in-game and have balanced server populations against performance.
Kilsin said:We have no idea Jacasta, it is impossible to even guess at this point in time, that is what testing will be used for, to gather the data and see what pop-to-server performance ratio is best and then we will go from there.
How many servers will depend on how many people play and what number we deem to be suitable for each server, we want them to feel lived in and social without being over populated and have queues for everything.So this can't be answered until we get well into testing and have everything we need implemented in-game and have balanced server populations against performance.
Fair enough, can't say I'm too surprised by the response. Had a hunch that this would be the situation seeing that the game is still a bit early in progression. Will you guys have just one server for testing? (I would think so for pre-alpha but what about alpha and Beta?)
Along those lines:
Will servers have a population number next to it? This question was asked here https://www.pantheonmmo.com/content/forums/topic/4895/by-the-number but not officially answered.
Jacasta said:And based off of how many people first play, suppose it's a lot more than you guys expect and people can't play due to the servers being full, how quickly would you be able to get more servers up and running,
Kils,
Please correct me if I'm wrong but didn't Brad say once that the plan was to serve the game off the cloud (am I using the correct terminology here?) Thus making it pretty simple to add more "servers" if poulations became such that additional space was needed.
Unlike the old EQ days where they had to actually buy and set up a server.
@Zorkon yes indeed cloud based, and it is even in the FAQ 1.16.1
Jacasta said:Kilsin said:We have no idea Jacasta, it is impossible to even guess at this point in time, that is what testing will be used for, to gather the data and see what pop-to-server performance ratio is best and then we will go from there.
How many servers will depend on how many people play and what number we deem to be suitable for each server, we want them to feel lived in and social without being over populated and have queues for everything.So this can't be answered until we get well into testing and have everything we need implemented in-game and have balanced server populations against performance.
Fair enough, can't say I'm too surprised by the response. Had a hunch that this would be the situation seeing that the game is still a bit early in progression. Will you guys have just one server for testing? (I would think so for pre-alpha but what about alpha and Beta?)
It depends on numbers my friend, we will be testing the servers during testing as much as we will be testing the game, which is why there will be a lot of outages, crashes, down times etc. for our proper testing phases but it is just too early to predict, unfortunately.
Zorkon said:Jacasta said:And based off of how many people first play, suppose it's a lot more than you guys expect and people can't play due to the servers being full, how quickly would you be able to get more servers up and running,
Kils,
Please correct me if I'm wrong but didn't Brad say once that the plan was to serve the game off the cloud (am I using the correct terminology here?) Thus making it pretty simple to add more "servers" if poulations became such that additional space was needed.
Unlike the old EQ days where they had to actually buy and set up a server.
Yes, that is correct, they are cloud based and we can put one up nearly anywhere in the world in a short period of time. :)
Hannar said:I believe you, you know, Kilsin... but, I mean... let's just say for testing purposes, go ahead and open one up tonight somewhere in the South East USA and let's see how it works. Feel free to send me a link. Thx.
Hahaha, we better set one up just to be sure :P
Kilsin said:We have no idea Jacasta, it is impossible to even guess at this point in time, that is what testing will be used for, to gather the data and see what pop-to-server performance ratio is best and then we will go from there.
How many servers will depend on how many people play and what number we deem to be suitable for each server, we want them to feel lived in and social without being over populated and have queues for everything.So this can't be answered until we get well into testing and have everything we need implemented in-game and have balanced server populations against performance.
I know it is going to be tough one to answer, but ill give it a shot. What was the real purpose of having different threads only to different level of packages? It really seems rather difficult to get everyone on one page.
Pledges i would have gone a different way, but there is only so much you can give.
I've seen threads being talked about in champion threads, which cross references, and used to get shut down here. Kilsin you been nice to allow us to talk about it here too, but that seems too much work, on the mods and devs part.
Honestly? give them a pen with Pantheon on it, it would seem more fitting HAHA jking.
just saying is all, you can shrug all you want :)
Kilsin said:Hannar said:I believe you, you know, Kilsin... but, I mean... let's just say for testing purposes, go ahead and open one up tonight somewhere in the South East USA and let's see how it works. Feel free to send me a link. Thx.
Hahaha, we better set one up just to be sure :P
I know that you have stated that you wont be able to have an idea of servers prior to testing. My question is, has there been any thought or discussion about once you have a general idea of population needs, announcing the server names, or even placeholders (1, 2, 3, 4) so that guilds and players can get an idea of where other people are planning on going. Some people prefer more (or even less) populated servers.
Kilsin said:Zorkon said:Jacasta said:And based off of how many people first play, suppose it's a lot more than you guys expect and people can't play due to the servers being full, how quickly would you be able to get more servers up and running,
Kils,
Please correct me if I'm wrong but didn't Brad say once that the plan was to serve the game off the cloud (am I using the correct terminology here?) Thus making it pretty simple to add more "servers" if poulations became such that additional space was needed.
Unlike the old EQ days where they had to actually buy and set up a server.
Yes, that is correct, they are cloud based and we can put one up nearly anywhere in the world in a short period of time. :)
This isn't directed at you Kilsin, but I am going to quote your post because of your aanswer.
So that we're all clear. "In the cloud" really is just a marketing phrase that means "hosted on someone else's computers". Back in the day it was called "managed hosting".
Krixus said:7!
I mean honestly, the game is probably 2 years out. The world isn't even built. Why do you think there'd even be an answer to this? The question is so far out of order.
What makes you think that it's two years out? Are you provy to some information that the rest of us aren't? In their latest video, they said that pre-alpha would be "coming soon". I use that in quotes due to the vagueness. I can see pre-alpha taking a few months, alpha taking a few months, and beta taking six months or so. That's far from two years.
Kilsin,
If you guys really need to hammer the snot out of your test servers to see what sort of load that they can handle, there is an application out there called Load Runner. It used to be owned by Hewlett Packard. Now that they have sold off their software division, I am not sure who owns it. For raw application load testing, there's not much better on the market. I do, however, suspect that it is pretty costly. Not only can it be set to just raw hammer the box, you can also build actions for it to performm as keyboard strokes and mouse clickies to simulate an actual user based activity.
Kalok said:Kilsin,
If you guys really need to hammer the snot out of your test servers to see what sort of load that they can handle, there is an application out there called Load Runner. It used to be owned by Hewlett Packard. Now that they have sold off their software division, I am not sure who owns it. For raw application load testing, there's not much better on the market. I do, however, suspect that it is pretty costly. Not only can it be set to just raw hammer the box, you can also build actions for it to performm as keyboard strokes and mouse clickies to simulate an actual user based activity.
They are already doing something similar. See this thread: https://www.pantheonmmo.com/content/forums/topic/6827/share-the-load
Bazgrim said:Kalok said:Kilsin,
If you guys really need to hammer the snot out of your test servers to see what sort of load that they can handle, there is an application out there called Load Runner. It used to be owned by Hewlett Packard. Now that they have sold off their software division, I am not sure who owns it. For raw application load testing, there's not much better on the market. I do, however, suspect that it is pretty costly. Not only can it be set to just raw hammer the box, you can also build actions for it to performm as keyboard strokes and mouse clickies to simulate an actual user based activity.
They are already doing something similar. See this thread: https://www.pantheonmmo.com/content/forums/topic/6827/share-the-load
That's good info, Bazgrim. Thanks for linking too it.
Bazgrim said:You're welcome!
I posted the Load Runner comment because I have worked for organizations that have used it in the past. That's all this software does is load testing. It's very good at it. To the point where if you don't set up your tests correctly you can DDoS the device(s) that you are testing against pretty easily.....hehehehe
I know you guys are on AWS, but I've been to some Azure demos and the ease of spinning up a server on any contenent, with regional availability, was really cool. If it works the same, you have some powerful tools at your hands. You could satisfy almost any world market if there is demand for Pantheon.
Jacasta said: ... Furthermore, how many can be online at a given time on one server? Idk if you'd have the answers to that quite yet, that might be something to ask when you guys have the servers pretty much set up, but just curiosity do you have an estimate? With SWTOR, I'd say that theres usually +500ish (1000's pushing it, probably a lot less) on at a single time throughout the day on one of the larger servers, such as Harb and EH (Most of the other servers are pretty pitiful though). But those servers, as well as all the other ones, are usually on the 'Light' status, and maybe occasionally the 'Standard' during their more active times, so based off of that I'd think that they'd be maxed out around 3000 perhaps? Would you guys be around that, is that your goal? Or would you perhaps be looking for less space?This is mostly just to VR thinking (Hell, it's been what? 2-3 years at least now? I'm sure this has crossed your minds and has been discussed tbh) but I also need to pass my time and I found myself wondering what VR has planned on this type of thing.
If they don't have several design goals decided on this subject by now, they are going to have a bad time. What do I mean by that?
The number of NPC's in a zone, the zone layout, respawn rates, drop rates, and the target player population in a zone are all factors driven by design goals. You don't just build a dungeon, put 1000 NPC's in it, and say enjoy. If you do, things like "The Deep" in Luclin occur, with all the socially toxic consequences.
A modern kernel can support over 10,000 concurrent connections. Up to 12 Million, in fact. Doesn't mean you should, but you could. So, operating system limits aren't really an issue any more in this one area.
In the past, they were, but not today. With that in mind, there are server side considerations. Context switching becomes a real issue in high volume, high connection environments where lowest real time latency is a goal. Multi core and threading can assist with this, but only to a point. If your context switch latency exceeds your RTT , (or even approaching 50%) go home, you've failed. The server lag will be so high, no-one will want to play.
The design of how a zone is created and maintained is paramount. If you run multiple game servers on a single virtual machine, or one process per NPC context switching becomes your nemesis. If you run a process per zone, and a thread per NPC, then you will experience synch, locking, and contention issues per process. These issues will arise once you go from 10 to 100, or 100 to 1000 or similar order of magnitude leaps in scope. 10 NPC's in a zone, everything is peachy. 100 NPC's in a zone, everything seems ok, but is a bit laggy... then you try 200 and everything goes off the rails. Some of the demonstrated zones could easily hold 1000 npc's, but SHOULD they?
Then there's pathfinding. Oh the joys of pathfinding. This is exacerbated immensely by the public design goal of no leash lengths. Even EQ1 has leash lengths, despite what people claim, they're just very long. Once a target in EQ1 get's a certain distance away, all the creatures will pathfind away from the target, and eventually return to their aggro point. This can be demonstrated with bard speed and any outdoor zone trivially, today. But when you get over 100 NPC's all pathfinding, and you need to generate location updates to however many people are within LOS or within the zone, that is an unpredictable beast of potential load. Server side CPU load as well as network load. Essentially, you're giving players the ability to produce unpredictable, significant, asynchronous load that impacts all other clients connected to that game server.
In classic EQ1, lower guk and solb were often so overhunted that a naked level 1 wood elf rogue could walk from the entrance to naggy's room (or from dead side to the king's room) without aggro of any kind. That's how well people knew respawn times, and NPC spawn locations. There would be so many camps it was a profoundly negative experience, for casual paying customers. Presuming Visionary Realms doesn't want to duplicate such a blunder, and with the available technological solutions/resources, the design goal comes down to: What social environment do they want to create?
Do you want 10 groups (60 people) in the zone? If respawn rate is 10 minutes, and Time-To-Kill is 1 minute. Then each group needs 10 mobs. If TTK is 30 seconds, each group needs 20 mobs. That's either 100 mobs or 200 mobs. Quite a difference. Mob density and population placement should then be driven by TTK & respawn rate design goals. So if you don't know these things before you start building zones, you will end up compromising on challenge. After the fact, you end up having to increase TTK by nerfing combat overall, or lowering respawn to artificially create urgency which has it's own snowball effect on the economy and progression rates. Similarly, if you don't do these things, then you end up with players sitting around on their asses with no mobs to kill because everyone kills everything too quickly. Guess which ones people like Chris Perkins end up being forced into choosing, historically? :) Why not choose another option? Because no-one wants to pay designers to go back and re-tune entire zones of content. It's a horrific time sink. It's far quicker and easier to do what makes players unhappy.
Add up all the zones, and you have a server total for population, bringing us back to the original question. For arguments sake, based on the current map, let's say there are only 22 zones. Probably a little low, but still. Let's say there are 100 NPC's in each zone that can be killed for loot and XP. 2200 targets total. If population was evenly distributed (which it never is) then that would mean if TTK is 1 minute, respawn is 10 minutes, and group size is 6... that means about 220 groups could be killing non-stop, or 1320 players total in a given game server. (using totally unrealistic numbers)
One year after launch, when more than 50% of the players are max level, what you want to know is, how many of those 22 zones are a challenge or desirable for max level players to be in? 4? That means you would have 400 available targets, but 660 players (or 110 groups) all trying to kill them. 110 groups need 10 mobs per group, or 1100 targets, but there are only 400. Social toxicity guaranteed.
Modern MMO's are nothing but spreadsheets, tables, and math behind the scenes, which can all be quantified and adjusted as necessary. Certain zones may have far more targets, far higher TTK, far lower respawn rates, if desired. But the goals can all be set ahead of time and worked towards logically, provided Visionary Realms has learned from the mistakes of history.
