Clustering development insight

Nostalrius official announcements

Re: Clustering development insight

by theemus » Thu Mar 03, 2016 7:11 pm

I convinced a friend from work to make a character on here. He gave up at about level 6 because of the spell/ability delay.
theemus
Private
Private
 

Re: Clustering development insight

by Mdmadam » Thu Mar 03, 2016 8:26 pm

Kaiji wrote:
If that happens then eventually maybe another realm will be needed somewhere down the line.

For now, clustering the current PvP server will relieve lag by spreading the present population over several servers. Surely you don't contest this?


I understand and I understand its potential to fix the lag but its still only a band-aid solution to a greater problem. There could be a 50k pop server (exaggeration) with <100 ms for generally every client connected, but would the server still be fun with no lag and no where to go without running into a crowd? Just thinking in the long-term and I'm predicting the pop will continue to rise for at least another year.
User avatar
Mdmadam
Sergeant Major
Sergeant Major
 

Re: Clustering development insight

by Kaiji » Thu Mar 03, 2016 10:33 pm

Mdmadam wrote:Just thinking in the long-term and I'm predicting the pop will continue to rise for at least another year.


Well obviously long term there might be a need for more servers, but you said that clustering "won't solve the problem".

I doubt it'll give everyone around the world 5 millisecond ping, but I'm certain it will be a drastic improvement.
Kaiji
Sergeant
Sergeant
 

Re: Clustering development insight

by Plask » Sat Mar 05, 2016 1:45 am

I remember when the server launched and the goal was to deliver a authentic vanilla experience.

Currently:
  • Double peak population than retail vanilla servers.
  • Huge delay on things like mailing, mounting etc to try keep the lagg down.
  • During peak, it laggs anyways
  • Lower draw distance
Tradeoff for 10k population. So I guess the goal is to have as populated server as possible. Yes there is a PvE server that is fairly active but majority wants PvP server. The population could easily have been divided on 2 servers and a queue enforced to make sure people don't flood one server.

Countless of time bandaid fixes has been done, as the dynamic draw distance reduction and delay priority etc. Say the lagg is fixed, nothing is really fixed in the long run, as the problem will come back again when the population increase again. This is in no way fault of the development team, it's just that vanilla was never meant to be played like this, so would be interesting to hear what's the plan for the main server. Personally I think the lagg is the biggest issue. Yeah, other servers don't have 10k, but some have decent population - and while they crash they certainly don't lagg on the same scale.

I guess what I'm trying to say is, does population hold that much importance?
Sidesprang wrote:Defcap is overrated at current state of the game.
Plask
Sergeant Major
Sergeant Major
 

Re: Clustering development insight

by Setup » Sat Mar 05, 2016 6:42 am

Plask wrote:I guess what I'm trying to say is, does population hold that much importance?


Yes. It's probably the single most important factor in a private realm. I say this as someone who has played on several private realms over the years. There are two core reasons why:

(1) Private realms come with extremely high rates of player turnover. Most top guilds routinely lose several players every month -- a much higher rate of attrition than you experienced in retail. On realms with a more modest population, the high rate of attrition also puts severe limits on guilds.

If the one or two guilds on your faction who play on your timezone don't need your class or are otherwise incompatible, both you and the guild are out of luck. If you're a guild leader, the low population means you have to accept more mediocre players in your ranks because you don't have the benefit of being able to kick the bad ones. It means recruitment is more difficult and time consuming, and your roster is weaker on average.

(2) Population is the best guarantee we have of the realm being here tomorrow. I've watched plenty of excellent realms, well scripted, promptly updated and fixed, get abandoned by their developers due to population problems. The longevity of Nost as a realm is directly proportional to its population.

If you don't want Nost to be one of the many private realms that randomly decides to shut down without a single day's warning (looking at you, Scriptcraft), then embrace the population. Realms like Feenix were TERRIBLE when it came to certain aspects, like stability and quality scripting, but they stay alive for years and years for the sole reason that they have a healthy population.

The population has a high cost, which is being addressed as best it can by the admins. But unless you want to increase the difficulty of staffing a raid, and increase the chances of the server dying at any moment's notice, there's really no sensible alternative.
Image
User avatar
Setup
Stone Guard
Stone Guard
 

Re: Clustering development insight

by Uzephi » Sat Mar 05, 2016 9:17 am

Plask wrote:I remember when the server launched and the goal was to deliver a authentic vanilla experience.

Currently:
  • Double peak population than retail vanilla servers.
  • Huge delay on things like mailing, mounting etc to try keep the lagg down.
  • During peak, it laggs anyways
  • Lower draw distance
Tradeoff for 10k population. So I guess the goal is to have as populated server as possible. Yes there is a PvE server that is fairly active but majority wants PvP server. The population could easily have been divided on 2 servers and a queue enforced to make sure people don't flood one server.

Countless of time bandaid fixes has been done, as the dynamic draw distance reduction and delay priority etc. Say the lagg is fixed, nothing is really fixed in the long run, as the problem will come back again when the population increase again. This is in no way fault of the development team, it's just that vanilla was never meant to be played like this, so would be interesting to hear what's the plan for the main server. Personally I think the lagg is the biggest issue. Yeah, other servers don't have 10k, but some have decent population - and while they crash they certainly don't lagg on the same scale.

I guess what I'm trying to say is, does population hold that much importance?



Actually this fix is the most blizzlike one yet for lag. After the launch of TBC (and the many server crashes) Blizzard implemented server clustering for the high population each realm was starting to get. Albiet this was older server hardware with less people, but I remember many times "Outland" or"Kalimdor" would crash on my server and I had to log onto an alt until that cluster node rebooted.

While a second server would be the best option with the high population of the PvP realm, out of those 13k, who would switch? Would there be a "Blizzlike" option to xfer your already made toon over to the new server? or would you have to reroll? I only remember the high pop server xfer option avail late TBC and not in Vanilla.
Uzephi
Stone Guard
Stone Guard
 

Re: Clustering development insight

by Plask » Sun Mar 06, 2016 12:06 pm

Uzephi wrote:
Plask wrote:I remember when the server launched and the goal was to deliver a authentic vanilla experience.

Currently:
  • Double peak population than retail vanilla servers.
  • Huge delay on things like mailing, mounting etc to try keep the lagg down.
  • During peak, it laggs anyways
  • Lower draw distance
Tradeoff for 10k population. So I guess the goal is to have as populated server as possible. Yes there is a PvE server that is fairly active but majority wants PvP server. The population could easily have been divided on 2 servers and a queue enforced to make sure people don't flood one server.

Countless of time bandaid fixes has been done, as the dynamic draw distance reduction and delay priority etc. Say the lagg is fixed, nothing is really fixed in the long run, as the problem will come back again when the population increase again. This is in no way fault of the development team, it's just that vanilla was never meant to be played like this, so would be interesting to hear what's the plan for the main server. Personally I think the lagg is the biggest issue. Yeah, other servers don't have 10k, but some have decent population - and while they crash they certainly don't lagg on the same scale.

I guess what I'm trying to say is, does population hold that much importance?



Actually this fix is the most blizzlike one yet for lag. After the launch of TBC (and the many server crashes) Blizzard implemented server clustering for the high population each realm was starting to get. Albiet this was older server hardware with less people, but I remember many times "Outland" or"Kalimdor" would crash on my server and I had to log onto an alt until that cluster node rebooted.

While a second server would be the best option with the high population of the PvP realm, out of those 13k, who would switch? Would there be a "Blizzlike" option to xfer your already made toon over to the new server? or would you have to reroll? I only remember the high pop server xfer option avail late TBC and not in Vanilla.

Transfer means transfer obviously, not making a new toon. And yeah, I think alot of players would rather play on a lagg free server, regardless of what is "blizzlike" or not. Besides, I don't see how blizzlike is a problem as there is already dynamic respawns and such due to the population.

Cluster nodes are cool, but I remember what happened when they came in WOTLK. Suddenly, server was holding way more players, but you only saw a fraction of them because being on different nodes. And most of the time you had to invite your friends to even see them. There's drawbacks to this method aswell although I do favor the positives. Not sayin' it's gonna be like that here but considering the high population I do feel like it's bound to happen. (And if it does, please put Chinese on a separate node).

I still don't understand whats wrong with having a normal server population, cluster nodes makes sense if you have a bunch of big servers and instead of opening 50 more servers you can increase capacity on your existing ones. However, in a small community like Nost I think it would make more sense to have Very High population servers (3-5k players) where there can form actual communities for each servers. This is the charm of vanilla.

Besides, they were talking about crossrealm BGs, so having another PvP server with same progression would make sense.
Sidesprang wrote:Defcap is overrated at current state of the game.
Plask
Sergeant Major
Sergeant Major
 

Re: Clustering development insight

by Niko » Sun Mar 06, 2016 12:20 pm

but how do we bait chinas into migrating server?
i would move, so would alot of others.
but then the chinamen would move along............
Last edited by Pottu on Fri Oct 30, 2015 5:31 pm, edited 3 times in total.
Reason: Put some trousers on
User avatar
Niko
Legionnaire
Legionnaire
 

Re: Clustering development insight

by RandyZie » Sun Mar 06, 2016 5:23 pm

Plask wrote:Cluster nodes are cool, but I remember what happened when they came in WOTLK. Suddenly, server was holding way more players, but you only saw a fraction of them because being on different nodes..


Please, read the damn thread. This will not happen. At most each zone could be split off from a continent (gate opening in silithus having server dedicated to it for example), but there will not be different instances of the same zone.
Priest
User avatar
RandyZie
Grunt
Grunt
 

Re: Clustering development insight

by Plask » Sun Mar 06, 2016 7:47 pm

RandyZie wrote:
Plask wrote:Cluster nodes are cool, but I remember what happened when they came in WOTLK. Suddenly, server was holding way more players, but you only saw a fraction of them because being on different nodes..


Please, read the damn thread. This will not happen. At most each zone could be split off from a continent (gate opening in silithus having server dedicated to it for example), but there will not be different instances of the same zone.

And what happens when the population increases to more then the current 13,5k record peak, and they want to increase draw distance back to original, and remove delays? The load is gonna increase and there is obviously gonna be compromises. Anyways as I said I think it's great, however I rather see them prioritize AQ, Naxx and TBC development. Do you remember the initial timeline for content at release and how many times it's been altered? Server was fine when the population was 5-6k, and tbh that was plenty enough.

Yes, I doubt they expected as big population as they have now but my point is having stable servers at 5k pop would have been more than acceptable and instead we could get some content and work could be done to prepare TBC.
Sidesprang wrote:Defcap is overrated at current state of the game.
Plask
Sergeant Major
Sergeant Major
 

PreviousNext

Return to Announcements