DIR-825 QOS Engine config

[hachi]

I have a small lan going through a dir-825 (rev.A1 1.10NA) and 15/2 cable connection, with a public host that runs dedicated game servers and an apache httpd. address allocation is both static/dhcp, with permanent boxes given their own ip while mobile devices grab one from the router in a different range.

I want to give the dedicated host priority for outbound traffic, but none of the settings I've tried so far seem to have any noticable effect on queueing. that is to say whenever I have any upstream saturating the connection elsewhere, with minimal inbound traffic, I still see latency for all connections to that host increase to about double the usual at that point, then backing down instantly as the line frees up.

am I doing this wrong or just expecting too much out of the qos engine? I thought I would see at least a relatively low increase with priority in action, but >100% the normal rate just doesn't seem very effective at all.

here are my settings:

WAN Traffic Shaping
Enable Traffic Shaping: on
Automatic Uplink Speed : off
Measured Uplink Speed : off
Manual Uplink Speed : 2048 kbps
Connection Type : cable/broadband
xDSL/Frame Relay: No

QoS Engine Setup
Enable QoS Engine : on
Automatic Classification : on
Dynamic Fragmentation : on

then for qos engine rules I have a single entry, with top priority pointing to the lan address of that host for all local/remote ports.

 

[thiggins]

You should look at the Active Sessions page to see the priorities that are actually being assigned. My guess is that you might have to manually enter a rule to dial down priority on applications that are competing for bandwidth with your server.

 

[hachi]

thanks for the suggestions, I did check the active sessions in the log, and everything from there looks as if it is doing what was intended. it shows all connections going to the designated lan address with top priority, and all outbound from others at lowest default.

but you are probably right in thinking there is some competition going on between applications, because from what I can see in the logs there is no priority listing for outbound connections coming from my prefferred address, and there are no application level settings. the problem here is that it appears the logic of this engine only pays attention to ranking up connections initiated by the wan side of the table. which of course is not going to do what I want, because the bulk of my preffered traffic is going to be session-less udp. so I'm guessing that these connections are getting the same default priority of 128 (out of possible 255) with the rest of the unclassified traffic.

faulty logic imo, since now the only solution I can see is giving the entire range of addresses other than my host a value of >128. I'm going to try this for now, if anyone can think of a better way to do this, by all means.

 

[hachi]

problem solved, for future reference:

apparently I had my manual uplink speed set too high, and wan throttle was interfering with the engine's outbound lan priority settings. my connection is 15/2, and speed tests put it well beyond the 2048kb/s upper limit I had it set for, but I guess it figures this speed is not stable.

so what I did was reduce it to a safer limit like 85% or 1748kb/s, so that wan throttle should never interfere with lan priorities. doing this in combination with public hosts set to priority of 1, and the rest of the range set to >129, the router was able to keep server latency completely stable.

 

[thiggins]

Thanks for reporting back. An incorrect uplink speed would definitely cause problems.