open All Channels
seplocked Features and Ideas Discussion
blankseplocked How to reduce fleet battle lag without having to add more servers
 
This thread is older than 90 days and has been locked due to inactivity.


 
Author Topic

Pulivin Motic
Caldari
Outer Limits Trading Company
Posted - 2010.01.16 04:58:00 - [1]
 

Pretty simple actually, have maby two employees monitor a real time model of 0.0 watching for spikes in the amount of players in the systems throughout. When a spike in players is detected they could simply monitor the rout that the large fleet is heading and look to see if there is another group of lag spikes leading to the same region then simply reinforce that area for as long as it takes the fighters to dissipate.
Neutral


Archeas
Posted - 2010.01.16 06:48:00 - [2]
 

What? What do you mean, "reinforce"? I didn't know servers had a magical "work harder" button on them. I'm not familiar with the eve clusters, but I'm sure a certain amount of load balancing automatically takes place, and I don't believe having a couple people constantly monitor things would help. Not only that, but the task of "monitoring" certain areas and systems in the eve universe is quite improbable, especially when it comes to trying to anticipate where a large group of players is headed.

Serge Bastana
Gallente
GWA Corp
Posted - 2010.01.16 07:13:00 - [3]
 

Originally by: Archeas
What? What do you mean, "reinforce"? I didn't know servers had a magical "work harder" button on them. I'm not familiar with the eve clusters, but I'm sure a certain amount of load balancing automatically takes place, and I don't believe having a couple people constantly monitor things would help. Not only that, but the task of "monitoring" certain areas and systems in the eve universe is quite improbable, especially when it comes to trying to anticipate where a large group of players is headed.


There's a message regarding the fleet notification form on the log in screen, filling this out prior to a big fleet fight allows CCP to allocate more server resources. So yes, there is a magical "work harder" button in essence.

Not sure if having CCP staff dedicated to monitoring such traffic would be a workable option, what could possibly be done is some form of alert system if systems within the same region spike above a given number each. Whether they could anticipate where two large fleets were going to meet I don't know. It might be a bit of a stretch asking them to.

darius mclever
Posted - 2010.01.16 07:17:00 - [4]
 

simple flaw in your plan:
reinforcing nodes is only possible at downtime as it stands.

Chris2746
Gallente
The Collective
Against ALL Authorities
Posted - 2010.01.16 08:04:00 - [5]
 

While this would not do anything for spontaneous battles that erupt, there should be certain triggers that automatically mark the system for at least a little bit of extra resources next time the servers go down. some of the things that should trigger this would be easily traceable with the new dominion expansion, and these would include things like on-lining an SBU, TCU, or ihub. or putting one of them into reinforced mode. as odds are more likely than not most groups will not let their territory go without a fight.

Also, I'm not sure if this is done already to some extent, but adaptive load distribution would be handy too, so say for example system A has been alliance X's base of operation for some time, and thus extra server resources have been allocated for it, if system A suddenly stops having a significant amount of player activity, and instead alliance X has moved their operations to system B, it should have some way of automatically adjusting the resources accordingly.

Savatar Mei
Posted - 2010.01.16 10:35:00 - [6]
 

you dont know whats causing the problems so u cant comment on a possible fix.

McEivalley
Cutting Edge Incorporated
RAZOR Alliance
Posted - 2010.01.16 11:24:00 - [7]
 

Originally by: Savatar Mei
you dont know whats causing the problems so u cant comment on a possible fix.


Actually he can, even if he doesn't know. Obviously, it doesn't mean that he is right - just that he can.

sheat posting ftw.

Germaldi's sister
Amarr
Posted - 2010.01.16 13:58:00 - [8]
 

problem with the 24 hr notice for server boost is u cant always predict when a battle is gonna occur.

Known warzones should be perma boosted imo


Future Mutant
Republic Military School
Posted - 2010.01.16 16:48:00 - [9]
 

Originally by: Germaldi's sister
problem with the 24 hr notice for server boost is u cant always predict when a battle is gonna occur.

Known warzones should be perma boosted imo




And if it was remotely feasible from a cost standpoint- it would have been done.

Niji Cura
Posted - 2010.01.18 03:42:00 - [10]
 

The problem with the direction of this thread is that there is no proper understanding of what exactly CCP does when it comes time to push the this magical "work harder" button on the server.

Let me start off with a brief insight on how the eve cluster works as detailed from one of the dev blogs I read from some time ago. The EVE cluster is basically made up of a whole bunch of servers. Each server runs it's own little simulation of TWO (2) specific solar systems. Systems that are known to have heavy traffic or usage are placed on beefier servers to handle the load. Jita for instance has a server all to itself (it still runs two sol systems, but the second is just a dummy system).

Now the magic behind the "work harder" button can be brought to light. When CCP gets those fleet tip offs, basically that allows them load the system in question onto a beefier server, or onto it's own dedicated server. That is my understand as to how it works, but to any dev out there, please correct me if I am wrong.

Now, I'd love it if the cluster itself would be able to balance it's resources on the fly (and I'm sure those in Virtual World Operations would love that to), and that's seems to be the intention for the creation of this thread. Unfortunately, the current setup of the EVE cluster would make this a daunting task. As of now, I don't believe there is any way to move an entire sol simulation for one server to another server while the cluster is up, probably because the engineering behind it would be very difficult. You'd need to figure out where to move one sol, load the sol to the destination server (probably the beefier server in this case), and then some how let the servers make a hand off of the simulation all while keeping data flowing to the clients still in the system. The process is even more difficult should there be a need two swap two sol simulations. Hopefully you get the point, that complexity is immense, and I don't see it happening for a long time.


 

This thread is older than 90 days and has been locked due to inactivity.


 


The new forums are live

Please adjust your bookmarks to https://forums.eveonline.com

These forums are archived and read-only