Connect Servers stability analysis

**Kirgston** · 26-03-11

Good day, everyone! Actually a few words from me personally (original text not write by me, i only translate and post this). First of all, all [BoR] Team must read this topic at breakfast, lunch and dinner. You guys wanted it and asked a long time... I also do not take upon itself any legal copyrights and author of the text - DoS.Ninja. A huge thank him for it .

Lets Start!

Almost every administrator MuOnline server that does not live in the tundra is aware of such a program as MCS (Mu Connect Server) from also little known [BoR] Team. This program is a "very good" release from the team (S@nek particular) and rapidly spread on the Internet including the foreign server. About old WebZen CS can say already forgotten, because it really was more demanding and less rapid than the analogous MCS. For all the time working with MCS, I often changed the version and there have been cases of failure of the program for which I was not paying attention, but recently I came across the apparent vulnerability of the program, the fact that the developer S@nek not well thought out program behavior at high loads . Even formed the impression that she just simply has not been tested under conditions such that even very bad. Well I decided to take the initiative in their hands, and tested the different versions of the program and compared them with counterparts from other manufacturers

Tested Programs:
- Mu Connect Server 1.0.4.360
- Mu Connect Server 1.0.7.300
- Mu Connect Server 1.0.8.950
- Mu Connect Server 1.0.9.200
- Mu Connect Server 1.0.9.530
- Mu Connect Server 1.0.9.750
- Mu Connect Server 1.0.9.1000
- Connect Server 1.0.6 External Client
- SCF Connect Server 1.0
- [P4F] Connect Server 0.2 (dev)
- Webzen Mu Connect Server 20050502

Tools:
- Written test emulator of a large number of connections and DOS attacks
- 2 computers, a server-client and client

Mu Connect Server 1.0.4.360 - version is not resistant to high loads
Test results:
Withstood 1,954 conn. then stop accept new connections

Mu Connect Server 1.0.7.300 - version is not resistant to high loads
Test results:
Withstood 1,940 conn. then stop accept new connections

Mu Connect Server 1.0.8.950 - this version is unstable even for low loads
Test results:
30 conn. - the program resets the connection and then you can reconnect
700 conn. - Led to the conclusion that errors in the logging window, then the server closed the connection port 44405 and did not accept more connections before you restart
When a small-time load program stop accept new connections until restarting, so you can easily remove it from the operating condition

Mu Connect Server 1.0.9.200 - is the same instability as in the 1.0.8.950
Test results:
20 conn. - the program resets the connection and then you can reconnect
700 conn. - stop accept new connections and close port 44405
Previous versions to reduce the limit of connections from 30 to 20. That is MCS stop accept new connections even earlier ...

Mu Connect Server 1.0.9.530 - the unstable version, the same problem with self-latching
Test results:
50 conn. - the program resets the connection and then you can reconnect
700 conn. - stop accept new connections and close port 44405
All of this is the limit already increased to 50 Connection

Mu Connect Server 1.0.9.750 - behavior is the same as in the 1.0.9.530

Mu Connect Server 1.0.9.1000 - absolutely not stable version
Test results:
50 conn. - drop the program in a second, the logs indicate an error:
[TTCPContext-> LoadExtensionFunction] ErrorCode: 10038

Connect Server 1.0.6 External Client – absolutely not stable version
Test results:
50 conn. - Program stop accept new connections at the time of connection and has caused a wild strain on the system. After disabling operability of the program survived. (This program is a kind of continuation of a series of MCS, it can be downloaded from the repository BoR Team. Above its development involved the whole team BoR + other people - approx. Killbrum)

[SCF] Connect Server 1.0 - stable version
Test results:
1000 conn. - Did not cause any interruption in the program and load on the system, the program stop accept new connections after 1000 connection and unlock after the test, while retaining the ability to connect to it

[P4F] Connect Server 0.2 - stable version
Test results:
1000 conn. - Did not cause any interruption in the program and the system load to a minimum, the program can easily stop accept new connections and released at the peak of connections.

Webzen Mu Connect Server 20050502 - stable version
Test results:
10000 conn. - I improvised means all still managed to emulate the 10,000 connections, and just judging by the same indicator is permissible up to 10 000 connections. Despite the burden the program's behavior was quite normal at the time of reaching 10,000 connections, the program itself blocked by maintaining the current connection, seen a small load on the system while disconnected from the connection server, in any case, after turning off CS efficiency persisted.

Well, having spent enough dough, I summarize:

Mu Connect Server [BoR] - do not have a stable version of the MCS from the team BoR! This program can be good in performance and functionality better than the original version of Webzen, but they are not resistant to high loads and DOS attacks. Version with support for IOCP (1.0.9 +) is generally better not to put it on your server, because judging by the tests, these versions are absolutely unstable and disabling low loads, which by the way, and not denied by the developer, just a manifestly unstable version and I'll take the product Connect Server 1.0.6 External Client which causes severe load on the system.

SCF Connect Server - the product passed the test perfectly, with no impact on the system program was able to adequately confront during exercise. But how this CS will behave in the SCF is not the server I find it difficult to answer because such tests, I still did not).

[P4F] Connect Server - given that the program is still under development, it shows remarkable results. Rate of speed and load on the system, even a little less than SCF.
Webzen Mu Connect Server - a good old original CS is a stable version, and in spite of his awkwardness coped with the test.

Instability [BoR] MCS
Yet why is this happening? I began to view logs of program development and i see what S@nek always make changes with anti-clogging ports, which he often finished the new versions, but he failed to properly implement it. As I understand it, the system has not been a normal debugging, or about just simply kept silent.

Credits: Dos.Ninja

Russian version:

Spoiler:

**MentaL**

**Faronnia** · 26-03-11

I always knew MCS was unstable, but these tests and technical info you posted are very much appreciated. Thanks man, great thread.

**nevS** · 26-03-11

nice info.
when im done with my .net connect server, could you test it too?

**Kirgston** · 03-04-11

Thx.

Yes sure i can test. Just send me PM with program =)

**nevS** · 18-04-11

Originally Posted by Kirgston

Thx.

Yes sure i can test. Just send me PM with program =)

im testing my server atm ;)
Getting 10000 connections is not a big problem.
How did you emulate the DOS-Attack? Which packets did u sent how often?

Thanks!

**-=FeliX=-** · 18-04-11

Originally Posted by Kirgston

Even formed the impression that she just simply has not been tested under conditions such that even very bad.

As I remember, S@nek is he!

Cya!

**Kirgston** · 18-04-11

sorry xD it make google translator xDD

**zohanel** · 01-07-11

But wat about the MTT Servers

**boncha** · 01-07-11

so what is the best CS > ?

**Bigman** · 01-07-11

igcDS is the best

**yannickisvet** · 09-07-11

Nice , GOnna check it soon!

**TRASPY** · 10-07-11

Originally Posted by Bigman

igcDS is the best

IGC ? Why ? Oh, because you are in this team

Rly, this is the best (IGCDS) ConnectServer but this work in IGC Files, and nothing more. Why? Because this is eMulator.

What is the best CS for many files ?

**Kirgston** · 10-07-11

In background only debuging [P4F] Launcher system , so look only to console (new [P4F] CS) and to task manager.

**Chern0byl** · 23-07-11

Definitely, P4F one is the best available so far.

**cswars** · 05-08-11

Originally Posted by Kirgston

Good day, everyone! Actually a few words from me personally (original text not write by me, i only translate and post this). First of all, all [BoR] Team must read this topic at breakfast, lunch and dinner. You guys wanted it and asked a long time... I also do not take upon itself any legal copyrights and author of the text - DoS.Ninja. A huge thank him for it .

Lets Start!

Almost every administrator MuOnline server that does not live in the tundra is aware of such a program as MCS (Mu Connect Server) from also little known [BoR] Team. This program is a "very good" release from the team (S@nek particular) and rapidly spread on the Internet including the foreign server. About old WebZen CS can say already forgotten, because it really was more demanding and less rapid than the analogous MCS. For all the time working with MCS, I often changed the version and there have been cases of failure of the program for which I was not paying attention, but recently I came across the apparent vulnerability of the program, the fact that the developer S@nek not well thought out program behavior at high loads . Even formed the impression that she just simply has not been tested under conditions such that even very bad. Well I decided to take the initiative in their hands, and tested the different versions of the program and compared them with counterparts from other manufacturers

Tested Programs:
- Mu Connect Server 1.0.4.360
- Mu Connect Server 1.0.7.300
- Mu Connect Server 1.0.8.950
- Mu Connect Server 1.0.9.200
- Mu Connect Server 1.0.9.530
- Mu Connect Server 1.0.9.750
- Mu Connect Server 1.0.9.1000
- Connect Server 1.0.6 External Client
- SCF Connect Server 1.0
- [P4F] Connect Server 0.2 (dev)
- Webzen Mu Connect Server 20050502

Tools:
- Written test emulator of a large number of connections and DOS attacks
- 2 computers, a server-client and client

Mu Connect Server 1.0.4.360 - version is not resistant to high loads
Test results:
Withstood 1,954 conn. then stop accept new connections

Mu Connect Server 1.0.7.300 - version is not resistant to high loads
Test results:
Withstood 1,940 conn. then stop accept new connections

Mu Connect Server 1.0.8.950 - this version is unstable even for low loads
Test results:
30 conn. - the program resets the connection and then you can reconnect
700 conn. - Led to the conclusion that errors in the logging window, then the server closed the connection port 44405 and did not accept more connections before you restart
When a small-time load program stop accept new connections until restarting, so you can easily remove it from the operating condition

Mu Connect Server 1.0.9.200 - is the same instability as in the 1.0.8.950
Test results:
20 conn. - the program resets the connection and then you can reconnect
700 conn. - stop accept new connections and close port 44405
Previous versions to reduce the limit of connections from 30 to 20. That is MCS stop accept new connections even earlier ...

Mu Connect Server 1.0.9.530 - the unstable version, the same problem with self-latching
Test results:
50 conn. - the program resets the connection and then you can reconnect
700 conn. - stop accept new connections and close port 44405
All of this is the limit already increased to 50 Connection

Mu Connect Server 1.0.9.750 - behavior is the same as in the 1.0.9.530

Mu Connect Server 1.0.9.1000 - absolutely not stable version
Test results:
50 conn. - drop the program in a second, the logs indicate an error:
[TTCPContext-> LoadExtensionFunction] ErrorCode: 10038

Connect Server 1.0.6 External Client – absolutely not stable version
Test results:
50 conn. - Program stop accept new connections at the time of connection and has caused a wild strain on the system. After disabling operability of the program survived. (This program is a kind of continuation of a series of MCS, it can be downloaded from the repository BoR Team. Above its development involved the whole team BoR + other people - approx. Killbrum)

[SCF] Connect Server 1.0 - stable version
Test results:
1000 conn. - Did not cause any interruption in the program and load on the system, the program stop accept new connections after 1000 connection and unlock after the test, while retaining the ability to connect to it

[P4F] Connect Server 0.2 - stable version
Test results:
1000 conn. - Did not cause any interruption in the program and the system load to a minimum, the program can easily stop accept new connections and released at the peak of connections.

Webzen Mu Connect Server 20050502 - stable version
Test results:
10000 conn. - I improvised means all still managed to emulate the 10,000 connections, and just judging by the same indicator is permissible up to 10 000 connections. Despite the burden the program's behavior was quite normal at the time of reaching 10,000 connections, the program itself blocked by maintaining the current connection, seen a small load on the system while disconnected from the connection server, in any case, after turning off CS efficiency persisted.

Well, having spent enough dough, I summarize:

Mu Connect Server [BoR] - do not have a stable version of the MCS from the team BoR! This program can be good in performance and functionality better than the original version of Webzen, but they are not resistant to high loads and DOS attacks. Version with support for IOCP (1.0.9 +) is generally better not to put it on your server, because judging by the tests, these versions are absolutely unstable and disabling low loads, which by the way, and not denied by the developer, just a manifestly unstable version and I'll take the product Connect Server 1.0.6 External Client which causes severe load on the system.

SCF Connect Server - the product passed the test perfectly, with no impact on the system program was able to adequately confront during exercise. But how this CS will behave in the SCF is not the server I find it difficult to answer because such tests, I still did not).

[P4F] Connect Server - given that the program is still under development, it shows remarkable results. Rate of speed and load on the system, even a little less than SCF.
Webzen Mu Connect Server - a good old original CS is a stable version, and in spite of his awkwardness coped with the test.

Instability [BoR] MCS
Yet why is this happening? I began to view logs of program development and i see what S@nek always make changes with anti-clogging ports, which he often finished the new versions, but he failed to properly implement it. As I understand it, the system has not been a normal debugging, or about just simply kept silent.

Credits: Dos.Ninja

Russian version:

Spoiler:

Man you good but i'm better.

Simple fix for all versions is:
If you host the game server you need FireDaemon program. Without this program
you will be lamer. You can configure this program and you can chose program
and time to restart, and automaticly restart if program have crash.
Ok, now we add Connect-Server and choose restart for every 15 minuts. It's all!

P.s.
You can use it for anyone programm!

Connect Servers stability analysis

Thread Tools

Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Re: Connect Servers stability analysis

Advertisement