Originally posted by MX13 Here are the possible reasons: 1) They truely believe the game is about to take off with new players, and they do not need Pre-NGE servers. Probability - Low 2) SOE can not support the code. Probability - Low 3) They truely believe the game is heading in the right dirrection, and believe the players will come back. Probability - Med 4) LA is not allowing them to run multiple game types. Probability - Med 5) By releasing Pre-NGE Servers, the flood of players to such servers would show their incompetence, costing many to loose their jobs or be personally tarnished for wasting Tens of Millions of dollars. Probability - High
NGE is taking off like one of those North Korea bottle rockets.
Originally posted by iskareot Originally posted by vbb1964 You can't run a game like this on 1 machine. It will take a cluster of over 40 machines just to get started. I think there will be issues but I think the movement is to get the financing to do this right up front.
I actually asked this question and it was 16 servers (however) understand that is because chat, i.e. is one, etc.. then they use different means to spread this out.
(They use profiles if you will) --
I did find it intresting that they use Unix -- for some things, like I do in my office. Hands down it is stable compared to Microcrap.
Thought a dev mentioned Linux at one point. Either way, I can see a few machines cobbled together being able to run the full game for a couple people but I can't imagine 1 SOE production server being any less than a 128 node cluster, probably more.
Until you cancel your subscription, you are only helping to continue the cycle of mediocrity.
Originally posted by boognish75 well for the love of god i hope they make one so we can get onto more interesting topics in the mmo world
LOL! Ya, I agree. It seems SWG is the most talked about MMO, even though it has relatively low subscriptions and so many people are angry about it.
As far as why SOE hasn't or won't release SWG Classic servers, like almost everyone else, I only have guesses based on what I've read. I would say that they really don't think it would be cost effective because they believe they can bring back the players who want Pre-Cu and would resubscribe if there were Classic servers by continuing to improve the NGE. I think they have written off many who cancelled and consider them to be lost, even if Classic servers were implemented. To start up and maintain even one Classic server may require a lot more money, time, computers, and technicians than we might think.
Thought a dev mentioned Linux at one point. Either way, I can see a few machines cobbled together being able to run the full game for a couple people but I can't imagine 1 SOE production server being any less than a 128 node cluster, probably more. A single machine should have no problem scaling to 100 or so users.
But at the end you get another bottleneck, the I/O one. Since each player must be updated about each other player in range, you get your I/O increasing with n^2 (where n is the number of players). This is only the network part, the number of packets and messages that need to be sent.
Another bottleneck will be the proximity checks (AoE weapons, range checks and collision detection).
Last is the database itself. While SWG is built on top of stored procedures and runs almost entirely within the database server itself (at least, that's my best guess), a database has a fixed capacity. Exceed it, and you're out of luck. That number occurs at a fixed point, based on total number of requests made and is closely tied to a certain number of active players.
Another major problem that is applicable to production servers (not emulators) is error handling. Getting a database connection timeout is simple case of reboot in emulators, but needs to be reliable handled on live servers. Similar problem is connection spikes during peak hours, during start up and during periods of high activity. During those times you get lots of lost/delayed packets which need to be re-sent or re-assembled, causing annoying delays that show across entire system.
Emulators can typically scale to several hundred players on a single machine, since they used highly simplified logic. Despite that, they all suffer from those problems, but almost always don't bother to address them, and simply let server crash or lag out if load becomes too big.
Production servers need to handle such situations gracefully, more often than not by sacrificing raw performance for reliability.
Comments
I actually asked this question and it was 16 servers (however) understand that is because chat, i.e. is one, etc.. then they use different means to spread this out.
(They use profiles if you will) --
I did find it intresting that they use Unix -- for some things, like I do in my office. Hands down it is stable compared to Microcrap.
Thought a dev mentioned Linux at one point. Either way, I can see a few machines cobbled together being able to run the full game for a couple people but I can't imagine 1 SOE production server being any less than a 128 node cluster, probably more.
Until you cancel your subscription, you are only helping to continue the cycle of mediocrity.
LOL! Ya, I agree. It seems SWG is the most talked about MMO, even though it has relatively low subscriptions and so many people are angry about it.
As far as why SOE hasn't or won't release SWG Classic servers, like almost everyone else, I only have guesses based on what I've read. I would say that they really don't think it would be cost effective because they believe they can bring back the players who want Pre-Cu and would resubscribe if there were Classic servers by continuing to improve the NGE. I think they have written off many who cancelled and consider them to be lost, even if Classic servers were implemented. To start up and maintain even one Classic server may require a lot more money, time, computers, and technicians than we might think.
A single machine should have no problem scaling to 100 or so users.
But at the end you get another bottleneck, the I/O one. Since each player must be updated about each other player in range, you get your I/O increasing with n^2 (where n is the number of players). This is only the network part, the number of packets and messages that need to be sent.
Another bottleneck will be the proximity checks (AoE weapons, range checks and collision detection).
Last is the database itself. While SWG is built on top of stored procedures and runs almost entirely within the database server itself (at least, that's my best guess), a database has a fixed capacity. Exceed it, and you're out of luck. That number occurs at a fixed point, based on total number of requests made and is closely tied to a certain number of active players.
Another major problem that is applicable to production servers (not emulators) is error handling. Getting a database connection timeout is simple case of reboot in emulators, but needs to be reliable handled on live servers. Similar problem is connection spikes during peak hours, during start up and during periods of high activity. During those times you get lots of lost/delayed packets which need to be re-sent or re-assembled, causing annoying delays that show across entire system.
Emulators can typically scale to several hundred players on a single machine, since they used highly simplified logic. Despite that, they all suffer from those problems, but almost always don't bother to address them, and simply let server crash or lag out if load becomes too big.
Production servers need to handle such situations gracefully, more often than not by sacrificing raw performance for reliability.