Re: [LARTC] AF

["Juergen" <lchou@staff.pccu.edu.tw>]

Sorry, I want to do your test but something wrong!!!!!!
tc qdisc add dev eth1 handle 1:0 root dsmark indices 64 set_tc_index


[root@altqtest linux-2.4]# tc qdisc add dev eth1 handle 1:0 root dsmark
indices 64 set_tc_index
Unknown qdisc "dsmark", hence option "indices" is unparsable

My kernel is 2.4.2 and iproute is 2.2.4-10
I have checked my kernel, and it support  field marker.
<M>   Diffserv field marker

Please give me some advice...
thx...

> # --- General setup  ---
> tc qdisc add dev eth1 handle 1:0 root dsmark indices 64 set_tc_index
> tc filter add dev eth1 parent 1:0 protocol ip prio 1 tcindex mask 0xfc
shift
> 2 pass_on
> tc qdisc add dev eth1 parent 1:0 handle 2:0 cbq bandwidth 10Mbit cell 8
> avpkt 1000 mpu 64
> tc filter add dev eth1 parent 2:0 protocol ip prio 1 tcindex mask 0xf0
shift
> 4 pass_on
>
> # --- AF Class 1 specific setup---
> tc class add dev eth1 parent 2:0 classid 2:1 cbq bandwidth 10Mbit rate
2Mbit
> avpkt 1000 prio 1 bounded allot 1514 weight 1 maxburst 21
> tc filter add dev eth1 parent 2:0 protocol ip prio 1 handle 1 tcindex
> classid 2:1
> tc qdisc add dev eth1 parent 2:1 gred setup DPs 3 default 2
>
> # --- AF Class 1 DP 1---
> tc filter add dev eth1 parent 1:0 protocol ip prio 1 handle 10 tcindex
> classid 1:111
> tc qdisc change dev eth1 parent 2:1 gred limit 60KB min 15KB max 45KB
burst
> 20 avpkt 1000 bandwidth 10Mbit DP 1 \
> probability 0.02 prio 2
>
> # --- AF Class 1 DP 2---
> tc filter add dev eth1 parent 1:0 protocol ip prio 1 handle 12 tcindex
> classid 1:112
> tc qdisc change dev eth1 parent 2:1 gred limit 60KB min 15KB max 45KB
burst
> 20 avpkt 1000 bandwidth 10Mbit DP 2 \
> probability 0.04 prio 3
>
> # --- AF Class 1 DP 3---
> tc filter add dev eth1 parent 1:0 protocol ip prio 1 handle 14 tcindex
> classid 1:113
> tc qdisc change dev eth1 parent 2:1 gred limit 60KB min 15KB max 45KB
burst
> 20 avpkt 1000 bandwidth 10Mbit DP 3 \
> probability 0.06 prio 4
>
> #------BE Queue setup------
> tc filter add dev eth1 parent 1:0 protocol ip prio 2 handle 0 tcindex mask
0
> classid 1:1
> tc class add dev eth1 parent 2:0 classid 2:5 cbq bandwidth 100Mbit rate
> 2Mbit avpkt 1000 prio 8 bounded allot 1514 weight 1 maxburst 21
> tc filter add dev eth1 parent 2:0 protocol ip prio 1 handle 0 tcindex
> classid 2:5
> tc qdisc add dev eth1 parent 2:5 red limit 60KB min 15KB max 45KB burst 20
> avpkt 1000 bandwidth 10Mbit probability 0.4
>
> ##########################
> Here are part of what i get when i "tc -s qdisc":
> qdisc red 8005: dev eth1 limit 60Kb min 15Kb max 45Kb
> Sent 0 bytes 0 pkts (dropped 0, overlimits 0)
>   marked 0 early 0qdisc gred 8001: dev eth1
> DP:1 (prio 8) Average Queue 0b Measured Queue 0b
> Packet drops: 0 (forced 0 early 0)
> Packet totals: 0 (bytes 0)  ewma 3 Plog 21 Scell_log 9
> DP:2 (prio 8) Average Queue 0b Measured Queue 0b
> Packet drops: 0 (forced 0 early 0)
> Packet totals: 0 (bytes 0)  ewma 3 Plog 20 Scell_log 9
> DP:3 (prio 8) Average Queue 0b Measured Queue 0b
> Packet drops: 0 (forced 0 early 0)
> Packet totals: 0 (bytes 0)  ewma 3 Plog 19 Scell_log 9
> Sent 0 bytes 0 pkts (dropped 0, overlimits 0)
>
> qdisc cbq 2: dev eth1 rate 10Mbit (bounded,isolated) prio no-transmit
> Sent 0 bytes 0 pkts (dropped 0, overlimits 0)
>   borrowed 0 overactions 0 avgidle 624 undertime 0
>
> qdisc dsmark 1: dev eth1 indices 0x0040 set_tc_index
> Sent 0 bytes 0 pkts (dropped 0, overlimits 0)
> pdrop 0 other 0
>
> #####################
>
> my question is why is the prio shown is always 8, although i set them to
> prio 2,3,4 respectively? And i suspect because of this, i actually get
> better performance for BE (in terms of lower delay).
>
> Also, i would like to clarify some requirement for af as proosed in RFC.
In
> "Appendix: Example services" of RFC 2597, it said that AF1x should have
> "greater probability for timely forwarding by having lighter load".
> Therefore, i send lighter traffic to class 1, than class 2 and 3. But
still
> i get lower delay for class BE, and no observable difference in delay for
> class 1, 2 and 3 (maybe it is the prio 8 problem as described above).
>
> What do you guys think about the delay for AF as proposed in RFC? Should
AF
> class 1 has the smallest delay compared to class 2,3 4 and BE?
>
> Last question, what do we do with excess AF traffic? Do we drop it or
> downgrade it to BE?
>
> If you have read to this part, i must thank you for your patience.
>
>
>
>
> _________________________________________________________________________
> Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
>
>
> _______________________________________________
> LARTC mailing list / LARTC@mailman.ds9a.nl
> http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO:
http://ds9a.nl/2.4Routing/
>


_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/2.4Routing/