PQ questions
Hi,
Quoted 35 lines
Hi all,
First, let me say I've been most impressed with how quickly and professionally people on this list ask and answer questions.
Next, let me say that with which I need help is properly configuring strict PQ, and gathering certain stats. Specifically:
- I need to create a priority queue with four queues (let's say they are of high, medium, normal, and low priority)
- I need to use tc filters such that:
- EF (
0xB8) goes to the high priority queue- AF21 (
0x50) goes to the medium priority queue- AF11 (
0x28) goes to the normal priority queue, and- BE traffic goes to the low priority queue
- For stat collection, I need to see:
- how many bytes and packets are in each of the four queues
- My configuration thus far is:
tc qdisc add dev
eml_testroot handle 1: prio bands 4 priomap 0 1 2 3tc filter add dev
eml_testparent 1:0 prio 1 protocol ip u32 match ip tos0xb80xffflowid 1:1tc filter add dev
eml_testparent 1:0 prio 2 protocol ip u32 match ip tos0x800xffflowid 1:2
Quoted 6 lines
tc filter add dev
eml_testparent 1:0 prio 3 protocol ip u32 match ip tos0x500xffflowid 1:3tc filter add dev
eml_testparent 1:0 prio 4 protocol ip u32 match ip tos0x000xffflowid 1:4
__________
Here is an article you may find useful:
http://citeseer.ist.psu.edu/539891.html
Here is the description of the configuration parameters of the PRIO qdisc:
http://www.lartc.org/howto/lartc.qdisc.classful.html#AEN903
(just in case you did not know what the "priomap" option is
used for)
Quoted 4 lines
My questions are:
- What if anything is missing/requiring change in my config given the stated requirements?
Your config does not prevent an higher priority class from starving a lower priority class. You can prevent it in two different ways (at least):
1) You can assign a TBF qdisc (Token Bucket) to the PRIO classes
TBF: http://www.lartc.org/howto/lartc.qdisc.classless.html#AEN691
2) You can replace the PRIO qdisc with something like HTB/CBQ
CBQ: http://www.lartc.org/howto/lartc.qdisc.classful.html#AEN939
HTB: http://luxik.cdi.cz/~devik/qos/htb/
Quoted 2 lines
- What if any command should I use to view how many bytes and packets are in each of the four queues?
The PRIO qdisc does not return statistics for its classes. However, a simple workaround consists of explicitly adding a qdisc to the four classes.
By default the PRIO qdisc assigns a pFIFO (packet FIFO) qdisc to its classes.
Here is how you can replace the 4 default pFIFO qdisc with 4 explicit pFIFO qdisc:
tc qdisc add dev eml_test parent 1:1 pfifo limit 1000 tc qdisc add dev eml_test parent 1:2 pfifo limit 1000 tc qdisc add dev eml_test parent 1:3 pfifo limit 1000 tc qdisc add dev eml_test parent 1:4 pfifo limit 1000
Now you can get the stats with:
tc -s -d qdisc list dev eml_test
Regards
/Christian
[ http://benve.info ]
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Hi Christian,
Thanks for the help. Please see my in-line comments:
Quoted 62 lines
-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Christian Benvenuti
Sent: Thursday, June 14, 2007 4:44 PM
To: [email protected]
Subject: [LARTC] Re: PQ questionsHi,
Quoted 6 lines
Hi all,
First, let me say I've been most impressed with how quickly and professionally people on this list ask and answer questions.
Next, let me say that with which I need help is properly configuring
strict
Quoted 3 lines
PQ, and gathering certain stats. Specifically:
- I need to create a priority queue with four queues (let's say they are
of
Quoted 26 lines
high, medium, normal, and low priority)
- I need to use tc filters such that:
- EF (
0xB8) goes to the high priority queue- AF21 (
0x50) goes to the medium priority queue- AF11 (
0x28) goes to the normal priority queue, and- BE traffic goes to the low priority queue
- For stat collection, I need to see:
- how many bytes and packets are in each of the four queues
- My configuration thus far is:
tc qdisc add dev
eml_testroot handle 1: prio bands 4 priomap 0 1 2 3tc filter add dev
eml_testparent 1:0 prio 1 protocol ip u32 match ip tos0xb80xffflowid 1:1tc filter add dev
eml_testparent 1:0 prio 2 protocol ip u32 match ip tos0x800xffflowid 1:2Quoted 6 lines
tc filter add dev
eml_testparent 1:0 prio 3 protocol ip u32 match ip tos0x500xffflowid 1:3tc filter add dev
eml_testparent 1:0 prio 4 protocol ip u32 match ip tos0x000xffflowid 1:4
__________Here is an article you may find useful:
http://citeseer.ist.psu.edu/539891.htmlHere is the description of the configuration parameters of the PRIO qdisc:
http://www.lartc.org/howto/lartc.qdisc.classful.html#AEN903
(just in case you did not know what the "priomap" option is
used for)
Quoted 8 lines
Quoted 3 lines
My questions are:
- What if anything is missing/requiring change in my config given the
stated
Quoted 1 line
requirements?
Your config does not prevent an higher priority class from starving a lower priority class.
Exactly. That is requirement.
Quoted 2 lines
You can prevent it in two different ways (at
least):
Don't want to prevent it right now.
Quoted 20 lines
1) You can assign a TBF qdisc (Token Bucket) to the PRIO classes
TBF: http://www.lartc.org/howto/lartc.qdisc.classless.html#AEN6912) You can replace the PRIO qdisc with something like HTB/CBQ
CBQ: http://www.lartc.org/howto/lartc.qdisc.classful.html#AEN939
HTB: http://luxik.cdi.cz/~devik/qos/htb/Quoted 1 line
- What if any command should I use to view how many bytes and packets are
in
Quoted 1 line
each of the four queues?
The PRIO qdisc does not return statistics for its classes. However, a simple workaround consists of explicitly adding a qdisc to the four classes.
By default the PRIO qdisc assigns a pFIFO (packet FIFO) qdisc to its classes.
Here is how you can replace the 4 default pFIFO qdisc with 4 explicit pFIFO qdisc:
Quoted 7 lines
tc qdisc add dev
eml_testparent 1:1 pfifo limit 1000 tc qdisc add deveml_testparent 1:2 pfifo limit 1000 tc qdisc add deveml_testparent 1:3 pfifo limit 1000 tc qdisc add deveml_testparent 1:4 pfifo limit 1000Now you can get the stats with:
tc -s -d qdisc list deveml_test
Those stats are nice to have, but the ones I must have are for how many bytes/packets are enqueued at whatever time I check the queues.
Quoted 5 lines
Regards
/Christian
[ http://benve.info ]
I have tried to configure PQ to have two queues per filter with no success. Is it even possible to have (what I'll call) hierarchical PQ? I have yet to find it.
Quoted 8 lines
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Hi,
Quoted 4 lines
Quoted 2 lines
Your config does not prevent an higher priority class from starving a lower priority class.
Exactly. That is requirement.
OK
Quoted 2 lines
Those stats are nice to have, but the ones I must have are for how many bytes/packets are enqueued at whatever time I check the queues.
That information is there. Here is an example:
(b=bytes p=packets)
#tc -s -d qdisc list dev eth1
qdisc prio 1: root bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
Sent 85357186 bytes 59299 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 35p requeues 0
+-> This field is not initialized for this
qdisc type
qdisc pfifo 10: parent 1:1 limit 1000p
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
rate 0bit 0pps backlog 0b 0p requeues 0
^^^^^^^^^^^^^
qdisc pfifo 20: parent 1:2 limit 1000p
Sent 85357120 bytes 59298 pkt (dropped 0, overlimits 0 requeues 0)
rate 0bit 0pps backlog 50470b 35p requeues 0
^^^^^^^^^^^^^^^^^^
qdisc pfifo 30: parent 1:3 limit 1000p
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
rate 0bit 0pps backlog 0b 0p requeues 0
^^^^^^^^^^^^^Quoted 1 line
I have tried to configure PQ to have two queues per filter with no success.
What do you mean?
Quoted 2 lines
Is it even possible to have (what I'll call) hierarchical PQ? I have yet to find it.
Something like this?
tc qdisc add dev eth1 handle 1: root prio tc qdisc add dev eth1 parent 1:1 handle 10 prio tc qdisc add dev eth1 parent 1:2 handle 20 prio tc qdisc add dev eth1 parent 1:3 handle 30 prio
Regards
/Christian
[ http://benve.info ]
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Slightly offtopic... Has anyone really experienced starving of low priority traffic with PRIO qdisc?
In my setup, I never achieved that, though I also wanted exactly that situation. I gave both the classes same amount of traffic at the same time. High prio got more bandwidth, but no starvation, even after I sent more traffic than the link capacity.
Quoted 2 lines
-----Original Message-----
From: [email protected]
[mailto:[email protected]]
Quoted 8 lines
On Behalf Of Christian Benvenuti
Sent: Friday, June 15, 2007 3:32 PM
To: [email protected]
Subject: [LARTC] Re: PQ questionsHi,
Quoted 1 line
Quoted 1 line
Your config does not prevent an higher priority class from
starving
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
a lower priority class.
Exactly. That is requirement.
OK
Quoted 1 line
Those stats are nice to have, but the ones I must have are for how
many
Quoted 25 lines
Quoted 1 line
bytes/packets are enqueued at whatever time I check the queues.
That information is there. Here is an example:
(b=bytes p=packets)#tc -s -d qdisc list dev eth1
qdisc prio 1: root bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
Sent 85357186 bytes 59299 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 35p requeues 0+-> This field is not initialized for this qdisc type qdisc pfifo 10: parent 1:1 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^ qdisc pfifo 20: parent 1:2 limit 1000p Sent 85357120 bytes 59298 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 50470b 35p requeues 0 ^^^^^^^^^^^^^^^^^^ qdisc pfifo 30: parent 1:3 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^
Quoted 6 lines
Quoted 1 line
I have tried to configure PQ to have two queues per filter with no
success.
What do you mean?
Quoted 1 line
Is it even possible to have (what I'll call) hierarchical PQ? I have
yet
Quoted 15 lines
to
Quoted 1 line
find it.
Something like this?
tc qdisc add dev eth1 handle 1: root prio tc qdisc add dev eth1 parent 1:1 handle 10 prio tc qdisc add dev eth1 parent 1:2 handle 20 prio tc qdisc add dev eth1 parent 1:3 handle 30 prio
Regards
/Christian
[ http://benve.info ]
Quoted 4 lines
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Hi,
a class is starved only if those with higher priority are
always (of pretty often) backlogged and do not give the lower priority classes a chance to transmit.
Therefore, if you transmit at a rate smaller than your CPU/s and NIC/s can handle you will not experience any starving.
For example, if you generate 50Mbit traffic on a 100Mbit NIC it is likely that you won't see any starving (unless your system is not able to handle 50Mbit traffic because of a complex TC or iptables configuration that consumes lot of CPU).
Regards
/Christian
[ http://benve.info ]
Quoted 52 lines · Salim S I
Slightly offtopic... Has anyone really experienced starving of low priority traffic with PRIO qdisc?
In my setup, I never achieved that, though I also wanted exactly that situation. I gave both the classes same amount of traffic at the same time. High prio got more bandwidth, but no starvation, even after I sent more traffic than the link capacity.Quoted 2 lines
-----Original Message-----
From: [email protected][mailto:[email protected]]
Quoted 8 lines
On Behalf Of Christian Benvenuti
Sent: Friday, June 15, 2007 3:32 PM
To: [email protected]
Subject: [LARTC] Re: PQ questionsHi,
Quoted 1 line
Quoted 1 line
Your config does not prevent an higher priority class from
starving
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
a lower priority class.
Exactly. That is requirement.
OK
Quoted 1 line
Those stats are nice to have, but the ones I must have are for how
many
Quoted 25 lines
Quoted 1 line
bytes/packets are enqueued at whatever time I check the queues.
That information is there. Here is an example:
(b=bytes p=packets)#tc -s -d qdisc list dev eth1
qdisc prio 1: root bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
Sent 85357186 bytes 59299 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 35p requeues 0+-> This field is not initialized for this qdisc type qdisc pfifo 10: parent 1:1 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^ qdisc pfifo 20: parent 1:2 limit 1000p Sent 85357120 bytes 59298 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 50470b 35p requeues 0 ^^^^^^^^^^^^^^^^^^ qdisc pfifo 30: parent 1:3 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^
Quoted 20 lines
Quoted 6 lines
Quoted 1 line
I have tried to configure PQ to have two queues per filter with no
success.
What do you mean?
Quoted 1 line
Is it even possible to have (what I'll call) hierarchical PQ? I have
yet
Quoted 13 lines
to
Quoted 1 line
find it.
Something like this?
tc qdisc add dev eth1 handle 1: root prio tc qdisc add dev eth1 parent 1:1 handle 10 prio tc qdisc add dev eth1 parent 1:2 handle 20 prio tc qdisc add dev eth1 parent 1:3 handle 30 prio
Regards
/Christian
[ http://benve.info ]
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
I tested on wireless link. It could give a maximum of 45Mbps. And I sent 30Mbps of both low prio and high prio traffic. Total of 60Mbps. My test was done with UDP, using tcpdump. When I increased the bandwidth to 40Mbps each, the high priority class got lesser bandwidth. (maybe the effect of the known issue that large amount of low prio traffic can starve high prio traffic)
Quoted 2 lines
-----Original Message-----
From: [email protected]
[mailto:[email protected]]
Quoted 17 lines
On Behalf Of Christian Benvenuti
Sent: Friday, June 15, 2007 4:16 PM
To: [email protected]
Subject: [LARTC] Re: PQ questionsHi,
a class is starved only if those with higher priority are
always (of pretty often) backlogged and do not give the lower priority classes a chance to transmit.
Therefore, if you transmit at a rate smaller than your CPU/s and NIC/s can handle you will not experience any starving.For example, if you generate 50Mbit traffic on a 100Mbit NIC it is likely that you won't see any starving (unless your system is not able to handle 50Mbit traffic because of a complex TC or iptables configuration that consumes lot of CPU).
Quoted 8 lines
Regards
/Christian
[ http://benve.info ]Quoted 3 lines · Salim S I
Slightly offtopic... Has anyone really experienced starving of low priority traffic with PRIO qdisc?
In my setup, I never achieved that, though I also wanted exactly
that
Quoted 1 line
Quoted 1 line
situation. I gave both the classes same amount of traffic at the
same
Quoted 1 line
Quoted 1 line
time. High prio got more bandwidth, but no starvation, even after I
sent
Quoted 13 lines
Quoted 13 lines
more traffic than the link capacity.
Quoted 2 lines
-----Original Message-----
From: [email protected][mailto:[email protected]]
Quoted 8 lines
On Behalf Of Christian Benvenuti
Sent: Friday, June 15, 2007 3:32 PM
To: [email protected]
Subject: [LARTC] Re: PQ questionsHi,
Quoted 1 line
Quoted 1 line
Your config does not prevent an higher priority class from
Quoted 8 lines
Quoted 8 lines
starving
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
a lower priority class.
Exactly. That is requirement.
OK
Quoted 1 line
Those stats are nice to have, but the ones I must have are for
how
Quoted 9 lines
Quoted 9 lines
many
Quoted 8 lines
Quoted 1 line
bytes/packets are enqueued at whatever time I check the queues.
That information is there. Here is an example:
(b=bytes p=packets)#tc -s -d qdisc list dev eth1
qdisc prio 1: root bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1
1
Quoted 1 line
Quoted 1 line
Quoted 1 line
Sent 85357186 bytes 59299 pkt (dropped 0, overlimits 0 requeues
0)
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
rate 0bit 0pps backlog 0b 35p requeues 0
+-> This field is not initialized for
this
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
qdisc type
qdisc pfifo 10: parent 1:1 limit 1000p
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
rate 0bit 0pps backlog 0b 0p requeues 0^^^^^^^^^^^^^qdisc pfifo 20: parent 1:2 limit 1000p
Sent 85357120 bytes 59298 pkt (dropped 0, overlimits 0 requeues
0)
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
rate 0bit 0pps backlog 50470b 35p requeues 0
^^^^^^^^^^^^^^^^^^ qdisc pfifo 30: parent 1:3 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^Quoted 1 line
I have tried to configure PQ to have two queues per filter with
no
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
success.
What do you mean?
Quoted 1 line
Is it even possible to have (what I'll call) hierarchical PQ? I
have
Quoted 16 lines
Quoted 14 lines
yet
Quoted 13 lines
to
Quoted 1 line
find it.
Something like this?
tc qdisc add dev eth1 handle 1: root prio tc qdisc add dev eth1 parent 1:1 handle 10 prio tc qdisc add dev eth1 parent 1:2 handle 20 prio tc qdisc add dev eth1 parent 1:3 handle 30 prio
Regards
/Christian
[ http://benve.info ]
Quoted 4 lines
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Hi Christian,
<snip>
Quoted 5 lines
#tc -s -d qdisc list dev eth1
qdisc prio 1: root bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
Sent 85357186 bytes 59299 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 35p requeues 0
Quoted 14 lines
+-> This field is not initialized for this
qdisc type
qdisc pfifo 10: parent 1:1 limit 1000p
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
rate 0bit 0pps backlog 0b 0p requeues 0
^^^^^^^^^^^^^
qdisc pfifo 20: parent 1:2 limit 1000p
Sent 85357120 bytes 59298 pkt (dropped 0, overlimits 0 requeues 0)
rate 0bit 0pps backlog 50470b 35p requeues 0
^^^^^^^^^^^^^^^^^^
qdisc pfifo 30: parent 1:3 limit 1000p
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
rate 0bit 0pps backlog 0b 0p requeues 0
^^^^^^^^^^^^^Yes, I can see that from your output. Here however is my config:
tc qdisc add dev eml_test root handle 1: prio bands 4 priomap 0 1 2 3
tc filter add dev eml_test parent 1:0 prio 1 protocol ip u32 match ip tos 0xb8 0xff flowid 1:1
tc filter add dev eml_test parent 1:0 prio 2 protocol ip u32 match ip tos 0x50 0xff flowid 1:2
tc filter add dev eml_test parent 1:0 prio 3 protocol ip u32 match ip tos 0x28 0xff flowid 1:3
tc filter add dev eml_test parent 1:0 prio 4 protocol ip u32 match ip tos 0x00 0xff flowid 1:4
tc qdisc add dev eml_test parent 1:1 handle 10: pfifo limit 2
tc qdisc add dev eml_test parent 1:2 handle 20: pfifo limit 2
tc qdisc add dev eml_test parent 1:3 handle 30: pfifo limit 2
tc qdisc add dev eml_test parent 1:4 handle 40: pfifo limit 2
___
Here is what I see when issuing the same command:
# tc -s -d qdisc list dev eml_test qdisc prio 1: bands 4 priomap 0 1 2 3 1 2 0 0 1 1 1 1 1 1 1 1 Sent 0 bytes 0 pkts (dropped 0, overlimits 0 requeues 0) qdisc pfifo 10: parent 1:1 limit 2p Sent 0 bytes 0 pkts (dropped 0, overlimits 0 requeues 0) qdisc pfifo 20: parent 1:2 limit 2p Sent 0 bytes 0 pkts (dropped 0, overlimits 0 requeues 0) qdisc pfifo 30: parent 1:3 limit 2p Sent 0 bytes 0 pkts (dropped 0, overlimits 0 requeues 0) qdisc pfifo 40: parent 1:4 limit 2p Sent 0 bytes 0 pkts (dropped 0, overlimits 0 requeues 0)
Quoted 5 lines
Quoted 2 lines
I have tried to configure PQ to have two queues per filter with no
success.
What do you mean?
Sorry, let me try to explain it this way (please refer to the above config):
- I presently have
-a strict PQ scheme which uses four queues
- four filters each of which determine what type of traffic gets
into which queue (EF, AF21, AF11 and BE respectively in my case)
- a specific pFIFO qdisc for each PQ "class"
__________
Quoted 11 lines
Quoted 2 lines
Is it even possible to have (what I'll call) hierarchical PQ? I have yet
to
Quoted 1 line
find it.
Something like this?
tc qdisc add dev eth1 handle 1: root prio tc qdisc add dev eth1 parent 1:1 handle 10 prio tc qdisc add dev eth1 parent 1:2 handle 20 prio tc qdisc add dev eth1 parent 1:3 handle 30 prio
(see above) I already have something just like this, just with pfifo for each child as opposed to the prio listed in the above config (thanks in great part to your previous help).
What I need is one more layer of hierarchy. Specifically, the queues defined by:
tc qdisc add dev eth1 parent 1:1 handle 10 prio
tc qdisc add dev eth1 parent 1:2 handle 20 prio
tc qdisc add dev eth1 parent 1:3 handle 30 prio
themselves need to be parents (e.g.):
tc qdisc add dev eth1 parent 10:0 handle 11 prio
tc qdisc add dev eth1 parent 20:0 handle 21 prio
tc qdisc add dev eth1 parent 30:0 handle 31 prio
Quoted 10 lines
Regards
/Christian
[ http://benve.info ]_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Please send me the exact config by which you got all those params in the output (especially backlog 0b 35p)... I just do not see that in mine.
Quoted 63 lines
-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Christian Benvenuti
Sent: Friday, June 15, 2007 3:32 AM
To: [email protected]
Subject: [LARTC] Re: PQ questionsHi,
Quoted 4 lines
Quoted 2 lines
Your config does not prevent an higher priority class from starving a lower priority class.
Exactly. That is requirement.
OK
Quoted 2 lines
Those stats are nice to have, but the ones I must have are for how many bytes/packets are enqueued at whatever time I check the queues.
That information is there. Here is an example:
(b=bytes p=packets)#tc -s -d qdisc list dev eth1
qdisc prio 1: root bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
Sent 85357186 bytes 59299 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 35p requeues 0+-> This field is not initialized for this qdisc type qdisc pfifo 10: parent 1:1 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^ qdisc pfifo 20: parent 1:2 limit 1000p Sent 85357120 bytes 59298 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 50470b 35p requeues 0 ^^^^^^^^^^^^^^^^^^ qdisc pfifo 30: parent 1:3 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^Quoted 1 line
I have tried to configure PQ to have two queues per filter with no
success.
What do you mean?
Quoted 1 line
Is it even possible to have (what I'll call) hierarchical PQ? I have yet
to
Quoted 1 line
find it.
Something like this?
tc qdisc add dev eth1 handle 1: root prio tc qdisc add dev eth1 parent 1:1 handle 10 prio tc qdisc add dev eth1 parent 1:2 handle 20 prio tc qdisc add dev eth1 parent 1:3 handle 30 prio
Regards
/Christian
[ http://benve.info ]
Quoted 4 lines
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Hi,
Quoted 2 lines · Salim S I
I tested on wireless link. It could give a maximum of 45Mbps. And I sent 30Mbps of both low prio and high prio traffic. Total of 60Mbps.
Do you mean to say that your wireless link can transmit at 45Mbps? If that's what you meant, what I meant to say is that if you generate almost (or more than) 45Mbps of high prio traffic than there is nothing (or almost nothing) left for the low prio traffic.
When you forward the above traffic (as opposed to when you generate it locally) there are other factors to take into account that can change the overall behavior.
For example, for each CPU there is one ingress queue that is shared by all ingress traffic that is received by interfaces whose driver does not use NAPI. These CPU queues are traversed before the ingress queueing disciplines and they have nothing to do with Traffic Control. It is possible therefore that under heavy load the low prio traffic fills in a significant portion of such CPU queues and reduces the amount of high prio traffic that reaches the egress queueing discipline (leaving therefore more chances to the low priority traffic to be scheduled).
Quoted 4 lines
My test was done with UDP, using tcpdump. When I increased the bandwidth to 40Mbps each, the high priority class got lesser bandwidth. (maybe the effect of the known issue that large amount of low prio traffic can starve high prio traffic)
Possible. See my comment above.
Regards
/Christian
[ http://benve.info ]
Quoted 85 lines
Quoted 2 lines
-----Original Message-----
From: [email protected][mailto:[email protected]]
Quoted 17 lines
On Behalf Of Christian Benvenuti
Sent: Friday, June 15, 2007 4:16 PM
To: [email protected]
Subject: [LARTC] Re: PQ questionsHi,
a class is starved only if those with higher priority are
always (of pretty often) backlogged and do not give the lower priority classes a chance to transmit.
Therefore, if you transmit at a rate smaller than your CPU/s and NIC/s can handle you will not experience any starving.For example, if you generate 50Mbit traffic on a 100Mbit NIC it is likely that you won't see any starving (unless your system is not able to handle 50Mbit traffic because of a complex TC or iptables configuration that consumes lot of CPU).
Quoted 8 lines
Regards
/Christian
[ http://benve.info ]Quoted 3 lines · Salim S I
Slightly offtopic... Has anyone really experienced starving of low priority traffic with PRIO qdisc?
In my setup, I never achieved that, though I also wanted exactlythat
Quoted 1 line
Quoted 1 line
situation. I gave both the classes same amount of traffic at the
same
Quoted 1 line
Quoted 1 line
time. High prio got more bandwidth, but no starvation, even after I
sent
Quoted 13 lines
Quoted 13 lines
more traffic than the link capacity.
Quoted 2 lines
-----Original Message-----
From: [email protected][mailto:[email protected]]
Quoted 8 lines
On Behalf Of Christian Benvenuti
Sent: Friday, June 15, 2007 3:32 PM
To: [email protected]
Subject: [LARTC] Re: PQ questionsHi,
Quoted 1 line
Quoted 1 line
Your config does not prevent an higher priority class from
Quoted 8 lines
Quoted 8 lines
starving
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
a lower priority class.
Exactly. That is requirement.
OK
Quoted 1 line
Those stats are nice to have, but the ones I must have are for
how
Quoted 9 lines
Quoted 9 lines
many
Quoted 8 lines
Quoted 1 line
bytes/packets are enqueued at whatever time I check the queues.
That information is there. Here is an example:
(b=bytes p=packets)#tc -s -d qdisc list dev eth1
qdisc prio 1: root bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1
1
Quoted 1 line
Quoted 1 line
Quoted 1 line
Sent 85357186 bytes 59299 pkt (dropped 0, overlimits 0 requeues
0)
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
rate 0bit 0pps backlog 0b 35p requeues 0
+-> This field is not initialized forthis
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
qdisc type
qdisc pfifo 10: parent 1:1 limit 1000p
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
rate 0bit 0pps backlog 0b 0p requeues 0^^^^^^^^^^^^^qdisc pfifo 20: parent 1:2 limit 1000p
Sent 85357120 bytes 59298 pkt (dropped 0, overlimits 0 requeues0)
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
rate 0bit 0pps backlog 50470b 35p requeues 0
^^^^^^^^^^^^^^^^^^ qdisc pfifo 30: parent 1:3 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^
Quoted 22 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
I have tried to configure PQ to have two queues per filter with
no
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
success.
What do you mean?
Quoted 1 line
Is it even possible to have (what I'll call) hierarchical PQ? I
have
Quoted 14 lines
Quoted 14 lines
yet
Quoted 13 lines
to
Quoted 1 line
find it.
Something like this?
tc qdisc add dev eth1 handle 1: root prio tc qdisc add dev eth1 parent 1:1 handle 10 prio tc qdisc add dev eth1 parent 1:2 handle 20 prio tc qdisc add dev eth1 parent 1:3 handle 30 prio
Regards
/Christian
[ http://benve.info ]
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Hi,
Quoted 2 lines · Tim Enos
Please send me the exact config by which you got all those params in the output (especially backlog 0b 35p)... I just do not see that in mine.
The configuration is the same as yours, with the difference that I have eth0 instead of eml_test.
I believe your config is OK.
I managed to get backlog!=0 by generating a huge amount of traffic with
mgen: 10K pkts/s of 1300bytes of size.
If you do not saturate your link it is likely you will not see anything sitting in the queue.
Regards
/Christian
[ http://benve.info ]
Quoted 58 lines
Quoted 58 lines
-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Christian Benvenuti
Sent: Friday, June 15, 2007 3:32 AM
To: [email protected]
Subject: [LARTC] Re: PQ questionsHi,
Quoted 4 lines
Quoted 2 lines
Your config does not prevent an higher priority class from starving a lower priority class.
Exactly. That is requirement.
OK
Quoted 2 lines
Those stats are nice to have, but the ones I must have are for how many bytes/packets are enqueued at whatever time I check the queues.
That information is there. Here is an example:
(b=bytes p=packets)#tc -s -d qdisc list dev eth1
qdisc prio 1: root bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
Sent 85357186 bytes 59299 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 35p requeues 0+-> This field is not initialized for this qdisc type qdisc pfifo 10: parent 1:1 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^ qdisc pfifo 20: parent 1:2 limit 1000p Sent 85357120 bytes 59298 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 50470b 35p requeues 0 ^^^^^^^^^^^^^^^^^^ qdisc pfifo 30: parent 1:3 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^Quoted 1 line
I have tried to configure PQ to have two queues per filter with no
success.
What do you mean?
Quoted 1 line
Is it even possible to have (what I'll call) hierarchical PQ? I have yet
to
Quoted 1 line
find it.
Something like this?
tc qdisc add dev eth1 handle 1: root prio tc qdisc add dev eth1 parent 1:1 handle 10 prio tc qdisc add dev eth1 parent 1:2 handle 20 prio tc qdisc add dev eth1 parent 1:3 handle 30 prio
Quoted 4 lines
Quoted 4 lines
Regards
/Christian
[ http://benve.info ]
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Cool,
Thanks Christian! I'm wishing that all of those same params showed up in the output without having to run anything. No problem. Should it matter that I'm using an emulated interface?
Also wondering what you think about my "hierarchical PQ" question. Have a good weekend.
Quoted 71 lines
-----Original Message-----
From: Christian Benvenuti [mailto:[email protected]]
Sent: Friday, June 15, 2007 2:57 PM
To: [email protected]
Cc: Tim Enos
Subject: RE: [LARTC] Re: PQ questionsHi,
Quoted 2 lines · Tim Enos
Please send me the exact config by which you got all those params in the output (especially backlog 0b 35p)... I just do not see that in mine.
The configuration is the same as yours, with the difference that I have eth0 instead of
eml_test.
I believe your config is OK.
I managed to get backlog!=0 by generating a huge amount of traffic with
mgen: 10K pkts/s of 1300bytes of size.
If you do not saturate your link it is likely you will not see anything sitting in the queue.Regards
/Christian
[ http://benve.info ]Quoted 2 lines
Quoted 2 lines
-----Original Message-----
From: [email protected] [mailto:lartc-Quoted 8 lines
Quoted 8 lines
On Behalf Of Christian Benvenuti
Sent: Friday, June 15, 2007 3:32 AM
To: [email protected]
Subject: [LARTC] Re: PQ questionsHi,
Quoted 1 line
Quoted 1 line
Your config does not prevent an higher priority class from
starving
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
a lower priority class.
Exactly. That is requirement.
OK
Quoted 1 line
Those stats are nice to have, but the ones I must have are for how
many
Quoted 25 lines
Quoted 25 lines
Quoted 1 line
bytes/packets are enqueued at whatever time I check the queues.
That information is there. Here is an example:
(b=bytes p=packets)#tc -s -d qdisc list dev eth1
qdisc prio 1: root bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
Sent 85357186 bytes 59299 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 35p requeues 0+-> This field is not initialized for this qdisc type qdisc pfifo 10: parent 1:1 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^ qdisc pfifo 20: parent 1:2 limit 1000p Sent 85357120 bytes 59298 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 50470b 35p requeues 0 ^^^^^^^^^^^^^^^^^^ qdisc pfifo 30: parent 1:3 limit 1000p Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) rate 0bit 0pps backlog 0b 0p requeues 0 ^^^^^^^^^^^^^
Quoted 21 lines
Quoted 6 lines
Quoted 6 lines
Quoted 1 line
I have tried to configure PQ to have two queues per filter with no
success.
What do you mean?
Quoted 1 line
Is it even possible to have (what I'll call) hierarchical PQ? I have
yet
Quoted 14 lines
Quoted 14 lines
to
Quoted 1 line
find it.
Something like this?
tc qdisc add dev eth1 handle 1: root prio tc qdisc add dev eth1 parent 1:1 handle 10 prio tc qdisc add dev eth1 parent 1:2 handle 20 prio tc qdisc add dev eth1 parent 1:3 handle 30 prio
Regards
/Christian
[ http://benve.info ]
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Quoted 4 lines · Salim S I
I tested on wireless link. It could give a maximum of 45Mbps. And I sent 30Mbps of both low prio and high prio traffic. Total of 60Mbps. My test was done with UDP, using tcpdump. When I increased the bandwidth to 40Mbps each, the high priority class got lesser bandwidth.
Maybe wireless is a special case here - was the driver/device actually on the prio box?
(maybe the
Quoted 2 lines
effect of the known issue that large amount of low prio traffic can starve high prio traffic)
On eth using tcp I can get prio to behave quite well.
You need to remember to filter arp to a high (best empty) class - it goes to x:2 by default, which made for a bit of wierdness when I tried last.
If you use tcp on my 100meg eth there is still a 300pkt buffer to fill before prio gets backlogged, so window scaling needs to be on and both ends need decent size buffers/scale amounts.
Maybe UDP would be different I'll have to try sometime.
Andy.
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Quoted 5 lines · Tim Enos
Cool,
Thanks Christian! I'm wishing that all of those same params showed up in the output without having to run anything. No problem. Should it matter that I'm using an emulated interface?
Quite possibly - using prio on real devices still can appear not to work until you have filled up any buffer the driver uses.
On my 100meg eth it would take 5/6 unscaled tcp connections to fill enough for prio to do anything.
You can use prio as a child of hfsc/htb so that they set the rate. It may be nicer to use htb's own prio though, if you need a slow rate and care about latency.
Andy.
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
It's PQ that is required. Here is what I have for config so far:
tc qdisc add dev eth0 root handle 1: prio bands 4 priomap 0 1 2 3
tc filter add dev eth0 parent 1:0 prio 1 protocol ip u32 match ip tos 0xb8 0xff flowid 1:1
tc filter add dev eth0 parent 1:0 prio 2 protocol ip u32 match ip tos 0x50 0xff flowid 1:2
tc filter add dev eth0 parent 1:0 prio 3 protocol ip u32 match ip tos 0x28 0xff flowid 1:3
tc filter add dev eth0 parent 1:0 prio 4 protocol ip u32 match ip tos 0x00 0xff flowid 1:4
tc qdisc add dev eth0 parent 1:1 handle 10: pfifo limit 2
tc qdisc add dev eth0 parent 1:2 handle 11: pfifo limit 2
tc qdisc add dev eth0 parent 1:3 handle 12: pfifo limit 2
tc qdisc add dev eth0 parent 1:4 handle 13: pfifo limit 2
__________
The above config works fine. The last four qdisc lines (handles 10: - 13:
inclusive) also work as prio if you leave out the 'limit' part of course.
The remaining part is to set children for the last four qdiscs (one for each). Said children qdiscs would have all the same attributes (as the parents (limit is something I'd change; the '2' is just an example). Is this possible?
Quoted 19 lines
-----Original Message-----
From: Andy Furniss [mailto:[email protected]]
Sent: Tuesday, June 19, 2007 6:17 PM
To: Tim Enos
Cc: 'Christian Benvenuti'; [email protected]
Subject: Re: [LARTC] Re: PQ questionsQuoted 3 lines · Tim Enos
Cool,
Thanks Christian! I'm wishing that all of those same params showed up in
the
Quoted 1 line
output without having to run anything. No problem. Should it matter that
I'm
Quoted 1 line
using an emulated interface?
Quite possibly - using prio on real devices still can appear not to work until you have filled up any buffer the driver uses.
Quoted 8 lines
On my 100meg eth it would take 5/6 unscaled tcp connections to fill enough for prio to do anything.
You can use prio as a child of hfsc/htb so that they set the rate. It may be nicer to use htb's own prio though, if you need a slow rate and care about latency.
Andy.
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Hi Tim, Andy,
Quoted 3 lines · Tim Enos
It's PQ that is required. Here is what I have for config so far:
tc qdisc add dev eth0 root handle 1: prio bands 4 priomap 0 1 2 3
Is "priomap 0 1 2 3" what you want/need or just a random mapping? (this is the default mapping that is used when none of the filters
matches)
Quoted 17 lines
tc filter add dev eth0 parent 1:0 prio 1 protocol ip u32 match ip tos
0xb80xffflowid 1:1tc filter add dev eth0 parent 1:0 prio 2 protocol ip u32 match ip tos
0x500xffflowid 1:2tc filter add dev eth0 parent 1:0 prio 3 protocol ip u32 match ip tos
0x280xffflowid 1:3tc filter add dev eth0 parent 1:0 prio 4 protocol ip u32 match ip tos
0x000xffflowid 1:4tc qdisc add dev eth0 parent 1:1 handle 10: pfifo limit 2
tc qdisc add dev eth0 parent 1:2 handle 11: pfifo limit 2
Quoted 8 lines
tc qdisc add dev eth0 parent 1:3 handle 12: pfifo limit 2
tc qdisc add dev eth0 parent 1:4 handle 13: pfifo limit 2
__________
The above config works fine. The last four qdisc lines (handles 10: - 13:
inclusive) also work as prio if you leave out the 'limit' part of course.
What do you mean?
Quoted 4 lines
The remaining part is to set children for the last four qdiscs (one for each). Said children qdiscs would have all the same attributes (as the parents (limit is something I'd change; the '2' is just an example). Is this possible?
Do you mean something like this?
tc qdisc add dev eth0 parent 10: handle 100: prio ... tc qdisc add dev eth0 parent 11: handle 110: prio ... tc qdisc add dev eth0 parent 12: handle 120: prio ... tc qdisc add dev eth0 parent 13: handle 130: prio ...
Why would you need to put a pfifo qdisc between the two prio qdisc? Wouldn't it be better to have
prio -> prio
OR
prio -> prio -> pfifo
instead of
prio -> pfifo -> prio ?
What criteria are you going to use to assign the right priority to the packets in the nested (i.e., 2nd level) prio qdisc?
Regards
/Christian
[ http://benve.info ]
Quoted 19 lines
Quoted 19 lines
-----Original Message-----
From: Andy Furniss [mailto:[email protected]]
Sent: Tuesday, June 19, 2007 6:17 PM
To: Tim Enos
Cc: 'Christian Benvenuti'; [email protected]
Subject: Re: [LARTC] Re: PQ questionsQuoted 3 lines · Tim Enos
Cool,
Thanks Christian! I'm wishing that all of those same params showed up in
the
Quoted 1 line
output without having to run anything. No problem. Should it matter that
I'm
Quoted 1 line
using an emulated interface?
Quite possibly - using prio on real devices still can appear not to work until you have filled up any buffer the driver uses.
Quoted 8 lines
Quoted 8 lines
On my 100meg eth it would take 5/6 unscaled tcp connections to fill enough for prio to do anything.
You can use prio as a child of hfsc/htb so that they set the rate. It may be nicer to use htb's own prio though, if you need a slow rate and care about latency.
Andy.
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Hi Christian,
Good morning, and thank you for proving me correct about how professional and responsive people on this list are (sincerely). Brief comments in-line:
Quoted 43 lines
-----Original Message-----
From: Christian Benvenuti [mailto:[email protected]]
Sent: Thursday, June 21, 2007 4:23 PM
To: Tim Enos
Cc: [email protected]; [email protected]
Subject: Re: PQ questionsHi Tim, Andy,
Quoted 3 lines · Tim Enos
It's PQ that is required. Here is what I have for config so far:
tc qdisc add dev eth0 root handle 1: prio bands 4 priomap 0 1 2 3
Is "priomap 0 1 2 3" what you want/need or just a random mapping? (this is the default mapping that is used when none of the filters
matches)Quoted 1 line
tc filter add dev eth0 parent 1:0 prio 1 protocol ip u32 match ip tos
0xb8Quoted 3 lines
0xffflowid 1:1tc filter add dev eth0 parent 1:0 prio 2 protocol ip u32 match ip tos
0x50Quoted 3 lines
0xffflowid 1:2tc filter add dev eth0 parent 1:0 prio 3 protocol ip u32 match ip tos
0x28Quoted 3 lines
0xffflowid 1:3tc filter add dev eth0 parent 1:0 prio 4 protocol ip u32 match ip tos
0x00Quoted 11 lines
0xffflowid 1:4tc qdisc add dev eth0 parent 1:1 handle 10: pfifo limit 2
tc qdisc add dev eth0 parent 1:2 handle 11: pfifo limit 2
tc qdisc add dev eth0 parent 1:3 handle 12: pfifo limit 2
tc qdisc add dev eth0 parent 1:4 handle 13: pfifo limit 2
Quoted 8 lines
Quoted 3 lines
__________
The above config works fine. The last four qdisc lines (handles 10: -
13:
Quoted 1 line
inclusive) also work as prio if you leave out the 'limit' part of
course.
What do you mean?
I mean that when saying something like:
"qdisc add dev eth0 parent 1:1 handle 10: prio limit 2"
you will get the following error (at least I do):
" What is "limit"? Usage: ... prio bands NUMBER priomap P1 P2..."
Changing the line like so works (and no error messages are generated):
"qdisc add dev eth0 parent 1:1 handle 10: prio"
Quoted 5 lines
Quoted 4 lines
The remaining part is to set children for the last four qdiscs (one for each). Said children qdiscs would have all the same attributes (as the parents (limit is something I'd change; the '2' is just an example). Is
this
Quoted 8 lines
Quoted 1 line
possible?
Do you mean something like this?
tc qdisc add dev eth0 parent 10: handle 100: prio ... tc qdisc add dev eth0 parent 11: handle 110: prio ... tc qdisc add dev eth0 parent 12: handle 120: prio ... tc qdisc add dev eth0 parent 13: handle 130: prio ...
Yes.
Quoted 8 lines
Why would you need to put a pfifo qdisc between the two prio qdisc? Wouldn't it be better to have
prio -> prioOR
Quoted 8 lines
prio -> prio -> pfifoinstead of
prio -> pfifo -> prio ?What criteria are you going to use to assign the right priority to the packets in the nested (i.e., 2nd level) prio qdisc?
The idea is that within each of the four priority classes/queues there would be two queues: one of some very small length (say 2) and another of some larger length (whatever the default is). So the thinking is that the traffic (having been marked by the application say) hits the top-level queue. If the traffic is marked EF, it will go into the highest priority queue. Once in that queue, it will hit the first pfifo (which in this model is 2 packets long). It will then hit the second pfifo queue before heading out onto the wire.
The ultimate concern is to know how many packets are in each of the priority queues at any given time.
Quoted 19 lines
Regards
/Christian
[ http://benve.info ]Quoted 11 lines
Quoted 11 lines
-----Original Message-----
From: Andy Furniss [mailto:[email protected]]
Sent: Tuesday, June 19, 2007 6:17 PM
To: Tim Enos
Cc: 'Christian Benvenuti'; [email protected]
Subject: Re: [LARTC] Re: PQ questionsQuoted 3 lines · Tim Enos
Cool,
Thanks Christian! I'm wishing that all of those same params showed
up in
Quoted 18 lines
Quoted 2 lines
Quoted 2 lines
the
Quoted 1 line
output without having to run anything. No problem. Should it matter
that
Quoted 4 lines
Quoted 4 lines
I'm
Quoted 1 line
using an emulated interface?
Quite possibly - using prio on real devices still can appear not to
work
Quoted 10 lines
Quoted 10 lines
until you have filled up any buffer the driver uses.
On my 100meg eth it would take 5/6 unscaled tcp connections to fill enough for prio to do anything.
You can use prio as a child of hfsc/htb so that they set the rate. It may be nicer to use htb's own prio though, if you need a slow rate and care about latency.
Andy.
_______________________________________________
LARTC mailing list
[email protected]
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc