13 replies to this topic

[blindman]

I have noticed with 0.33 that it generates a large amount of internal lan traffic, this is also true of nit and to som extent nq, so far jay is the only one I use that doesnt generate that much lan traffic.

Roughly 6 Mbps without bots, roughly 10 to 11 Mbps with 16 bots.
It would be a big improvement if it come down some on my network as I run servers from my home and have been for 6 years.

[gaoesa]

How many real players? The reason for my asking is that all the traffic happens between the server and the players. With the exception of PunkBuster, getstatus and getinfo server querys. How was the 10 to 11 Mbps calculated?

[blindman]

well there maybe three guys on at one point, i use a wnr4000 router loaded with dd wrt, theres a bandwidth graph as one of the additions, and I can say for sure that its way higher with silent, nit and nq.

[gaoesa]

Is the PunkBuster disabled or fully updated on the server and on the clients? One player has bandwidth usage that fluctuates between 2 and 8 KB/s both directions with server having 12 players. Averaging less then 5 kB/s. This was measured using netlimiter locally while on a normal streaming server and PunkBuster enabled.

6 mbit/s is about 732 KB/s and that is not normal for 3 clients. Bot's do not have direct influence to the bandwidth because they are integrated to the server. They do increase the amount of game events that are sent to the 3 players. However, influence of 4 mbit/s, sounds unbelievable.

EDIT:
It is quite possible the usage is higher then in Jaymod. The silEnT mod has features that Jaymod does not have. Try disabling g_spreeOptions and see if that has an effect. At least multikills is a thing that is not distributed to all clients in the Jaymod.

I'm thinking higher values maybe possible also because of the point where it is measured. The netlimiter, as far as I know, will display the actually used bandwidth for the data. When I would imagine, the router will be counting IP frames and more of the stuff that relates to the transmissions and not only the data.

[blindman]

pb is normally disabled. What i am really getting at here is why, I used to run 3 three servers all the time, two jaymod and one nq as little as 6 months ago with no issues. I''m not criticising, just wondering if with all the new improvements something may have got overlooked.
Now granted all my machines are not all new, but they should be fine for running et servers in linux.

I should point out that even jaymod used with enhmod causes more or less the same result, increased lan traffic that seems pointless, I have changed nic cards, subnets and even swapped machines, if it's a lan issue on my end im not seeing what else could be generating it.

Is the PunkBuster disabled or fully updated on the server and on the clients? One player has bandwidth usage that fluctuates between 2 and 8 KB/s both directions with server having 12 players. Averaging less then 5 kB/s. This was measured using netlimiter locally while on a normal streaming server and PunkBuster enabled.

6 mbit/s is about 732 KB/s and that is not normal for 3 clients. Bot's do not have direct influence to the bandwidth because they are integrated to the server. They do increase the amount of game events that are sent to the 3 players. However, influence of 4 mbit/s, sounds unbelievable.

EDIT:
It is quite possible the usage is higher then in Jaymod. The silEnT mod has features that Jaymod does not have. Try disabling g_spreeOptions and see if that has an effect. At least multikills is a thing that is not distributed to all clients in the Jaymod.

I'm thinking higher values maybe possible also because of the point where it is measured. The netlimiter, as far as I know, will display the actually used bandwidth for the data. When I would imagine, the router will be counting IP frames and more of the stuff that relates to the transmissions and not only the data.

im having a hard time with that, why would there be so much variance between mods.
If I understand netlimiter, it allows you to create rules for various programs running in the network, that doesnt seem to be my issue, the mod is just generating more traffic than it needs to.

Edited by blindman, 24 September 2011 - 04:33 PM.

[gaoesa]

My version of netlimiter only displays the transmission speed for each process. It's not commercial version which maybe has that kind of features.

We have decreased bandwidth in some areas. In other places it could have been slightly increased. Namely the g_spreeOptions. Overall, I believe it is significantly reduced from the ETPub. Significantly is a relative term because majority of the traffic is not related to mod features unless something really big is made. However, 6 mbit/s second is still surreal for 3 players. Since you are working this issue on so low level (far from the appication layer), you possibly should check QoS and other things that influence the traffic. For example dropping the network packets will cause retransmissions handled by the game. For network engineering, I'm not the person who has knownledge to help. I will fo some comparisons between mods at some point. I tried to test my old Jaymod 2.1.7 on my server but apparently I have allready deleted the bots so I couldn't create same bot conditions.

EDIT:
One more thing I must ensure you have set is the delta compression. Without this, the traffic will increase in huge amounts. The delta compression ensures that only changing information is sent to the client. The delta compression is per client setting so you should probably force the setting.

[blindman]

delta compression...how is that done?
I have done everything i caan think of that might cause it, nothing changes, Ive taken switches offline, changed cables etc, its actually worse with nitmod, but I cant even get to their forums, much less report on it.

Edited by blindman, 24 September 2011 - 05:14 PM.

[gaoesa]

Forcing delta compression to all clients happen by adding default.cfg to the mapconfigs directory. http://mygamingtalk....tml#forcesvcvar There should be a file for it in the download allready. Then, make sure it has line "sv_cvar cl_nodelta EQ 0" or alternatively "forcecvar cl_nodelta 0" these will force the clients to use that.

Some server settings that you can adjust for reducing network use in the mod:

g_landmineNotifyType 1 (http://mygamingtalk....dmineNotifyType)

This will send only 1 summary of all the revealed landmines. Option 3 will increase the usage a tiny bit.

g_spreeOptions (http://mygamingtalk....#g_spreeOptions) Disabling all the messages that are not needed will reduce the bandwidth needed.

g_misc 16 (http://mygamingtalk....ver.html#g_misc) Disabling announcing of the killer HP. This has a very minimal effect.

[blindman]

I will apply these settings.

[blindman]

There is some minor savings there, but it is small.

[blindman]

Hmm. seems if I change ports, all the idiots hammering my router go away, I know how this sounds, but I've had issues with ddos crap in the past. Wasn't enough to stop the servers it seems, I had to physically change ports on the firewall. So it seems the mod isn't at fault here, thanks for answering.

[gaoesa]

Glad to hear you found the solution.

[blindman]

M ost likely will happen again, isnt the first time I've had to do that.

[hellreturn]

M ost likely will happen again, isnt the first time I've had to do that.

Probably getstatus DDoS issue. I would suggest to try the gaoesa IPTable chain or schnoog script for stopping those attack.