Skip to contentSkip to conversation

Testing HTB with 2.5.68 Kernel.

LARTC Yaron Benita 2 participants 4 messages May 2, 2003 – May 9, 2003

lore
Share & tools
lore
#0 lore raw

Did anyone tried to run the HTB qdisc using 2.5.68
Kernel ?

I tried to test it, but it doesn't look like it is
working well. The same tests works using 2.3 and 2.4.20 kernels. Is there a new patch for 2.5 kernels ?

Yaron

__________________________________
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.
http://search.yahoo.com
_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

#1 lore raw
Quoted 7 lines · Yaron Benita

Did anyone tried to run the HTB qdisc using 2.5.68
Kernel ?

I tried to test it, but it doesn't look like it is
working well. The same tests works using 2.3 and 2.4.20 kernels. Is there a new patch for 2.5 kernels ?

I have just used wondershaper 1.1a on a 2.5.68 kernel, and in my limited testing everything is working as expected: packets are classified correctly, and shaping seems to work fine here.

By the way, what makes you thing it is not working ok on your setup ?. Some more detailled information would be helpful.

Signature · 5 lines
Jose Luis Domingo Lopez
Linux Registered User #189436     Debian Linux Sid (Linux 2.5.68)
_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
#2 lore raw
In reply to Yaron Benita
Quoted 5 lines · Yaron Benita

I enabled the HTB scheduler option in the kernel.
Than I created an HTB qdisc on eth1 using "tc" tool. I added a class with a rate limit of 30000kbits, and a filter attached to this class.

Please send your mails first to the mailing list, and maybe even to certain people. You will get more and better answers, and everything will get "archived" and be searcheable through usual search engines.

Quoted 8 lines

I used a traffic generator to send data to eth0 which
forwared the data to eth1 and than back to the traffic generator. I sent a bandwidth of 70000kbits, and the same traffic was forwarded through interface eth1.

This test shows that the traffic was not shaped to
30000kbits. The same test works greate in 2.4.20.

I have set up wondershaper 1.1a in my box. Configured it for an upload limit of 1000 Kbps, and used "netcat" on the server and client sides, measuring bandwidth with both "iptraf" and "gkrellm", and transfer rates are the ones configured.

The configuration used in my test follows. Traffic was generated on the client side via "cat /large/archive | nc remote_ip 8000", and received on the server side with "nc -l -p 8000 > /dev/null". Hope it helps.

#!/bin/bash
DOWNLINK00
UPLINK00
DEV=eth0
if [ "$1" = "status" ]
then
tc -s qdisc ls dev $DEV
tc -s class ls dev $DEV
exit
fi

# clean existing down- and uplink qdiscs, hide errors
tc qdisc del dev $DEV root    2> /dev/null > /dev/null
tc qdisc del dev $DEV ingress 2> /dev/null > /dev/null

if [ "$1" = "stop" ]
then
exit
fi

###### uplink

# install root HTB, point default traffic to 1:20:
tc qdisc add dev $DEV root handle 1: htb default 20
# shape everything at $UPLINK speed - this prevents huge queues in your
# DSL modem which destroy latency:
tc class add dev $DEV parent 1: classid 1:1 htb rate ${UPLINK}kbit burst 6k
# high prio class 1:10:
tc class add dev $DEV parent 1:1 classid 1:10 htb rate ${UPLINK}kbit \
   burst 6k prio 1
# bulk & default class 1:20 - gets slightly less traffic, 
# and a lower priority:
tc class add dev $DEV parent 1:1 classid 1:20 htb rate $[9*$UPLINK/10]kbit \
   burst 6k prio 2
tc class add dev $DEV parent 1:1 classid 1:30 htb rate $[8*$UPLINK/10]kbit \
   burst 6k prio 2
# all get Stochastic Fairness:
tc qdisc add dev $DEV parent 1:10 handle 10: sfq perturb 10
tc qdisc add dev $DEV parent 1:20 handle 20: sfq perturb 10
tc qdisc add dev $DEV parent 1:30 handle 30: sfq perturb 10
# TOS Minimum Delay (ssh, NOT scp) in 1:10:
tc filter add dev $DEV parent 1:0 protocol ip prio 10 u32 \
      match ip tos 0x10 0xff  flowid 1:10
# ICMP (ip protocol 1) in the interactive class 1:10 so we 
# can do measurements & impress our friends:
tc filter add dev $DEV parent 1:0 protocol ip prio 10 u32 \
        match ip protocol 1 0xff flowid 1:10
# To speed up downloads while an upload is going on, put ACK packets in
# the interactive class:
tc filter add dev $DEV parent 1: protocol ip prio 10 u32 \
   match ip protocol 6 0xff \
   match u8 0x05 0x0f at 0 \
   match u16 0x0000 0xffc0 at 2 \
   match u8 0x10 0xff at 33 \
   flowid 1:10
# rest is 'non-interactive' ie 'bulk' and ends up in 1:20

tc filter add dev $DEV parent 1: protocol ip prio 18 u32 \
match ip dst 0.0.0.0/0 flowid 1:20

########## downlink #############

# slow downloads down to somewhat less than the real speed  to prevent 
# queuing at our ISP. Tune to see how high you can set it.
# ISPs tend to have *huge* queues to make sure big downloads are fast
#
# attach ingress policer:

tc qdisc add dev $DEV handle ffff: ingress

# filter *everything* to it (0.0.0.0/0), drop everything that's
# coming in too fast:

tc filter add dev $DEV parent ffff: protocol ip prio 50 u32 match ip src \
0.0.0.0/0 police rate ${DOWNLINK}kbit burst 10k drop flowid :1

Signature · 5 lines
Jose Luis Domingo Lopez
Linux Registered User #189436     Debian Linux Sid (Linux 2.5.68)
_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
#3 lore raw
In reply to Yaron Benita

Hi,

I did a comprehensive testing with 2.3 and 2.4.20
kernels. I used an Ixia as a traffic generator for pumping data to the Linux
box. In the Linux box I have two interfaces, one is used for data input and
one is used for data output.

         -------            --------
         |     |1 ---> eth0 |Linux |
         |Ixia |2 <--- eth1 |      |
         -------            --------

I set the policing setting to eth1. The Linux box
forwarded every packet it
receives on interface eth0.

In my tests I tried different rate limits and
different packet sizes. I
found a linear relation between the burst and the
traffic, which give the
most accurate results.

I am trying to verify the same results using 2.5.68
kernel.

The reason it works for you is because you were using
a rate limit of
1000kbit and a packet size of 512. I verified your test, and it looks like
it works fine. When I tried different packet sizes, and different rates the
HTB scheduling doesn't work well.

The following is a sample of test results:
Test 1:
Pumping: 74Mbit/s
Rate limit: 1000Kbit

Results:

Packet size Received traffic (eth1)

      64             27Mbit/s
      128            42Mbit/s
      512            1.3Mbit/s
      1500           1.1Mbit/s

Test 2:
Pumping: 74Mbit/s
Rate limit: 2000Kbit

Results:

Packet size Received traffic (eth1)

      64             27Mbit/s
      512            4Mbit/s
      1500           2.4Mbit/s

Test 3:
Pumping: 74Mbit/s
Rate limit: 6000Kbit

Results:

Packet size Received traffic (eth1)

      64             27Mbit/s
      512            74Mbit/s
      1500           6Mbit/s

Test 4:
Pumping: 74Mbit/s
Rate limit: 8000Kbit

Results:

Packet size Received traffic (eth1)

      64             27Mbit/s
      512            74Mbit/s
      1500           12Mbit/s

As you can see from the results, the rate shaping
functionality doesn't
works in most
Cases. As I maintained before, these tests works
very well in 2.4.20.

Yaron

--- José_Luis_Domingo_López <[email protected]>

Quoted 136 lines · wrote:
Quoted 2 lines · Yaron Benita

I enabled the HTB scheduler option in the kernel.
Than I created an HTB qdisc on eth1 using "tc"

tool. I

Quoted 1 line

added a class with a rate limit of 30000kbits, and

a

Quoted 2 lines

filter attached to this class.

Please send your mails first to the mailing list, and maybe even to
certain people. You will get more and better answers, and everything
will get "archived" and be searcheable through usual search engines.

Quoted 1 line

I used a traffic generator to send data to eth0

which

Quoted 1 line

forwared the data to eth1 and than back to the

traffic

Quoted 1 line

generator. I sent a bandwidth of 70000kbits, and

the

Quoted 4 lines

same traffic was forwarded through interface eth1.

This test shows that the traffic was not shaped

to

Quoted 2 lines

30000kbits. The same test works greate in 2.4.20.

I have set up wondershaper 1.1a in my box. Configured it for an upload
limit of 1000 Kbps, and used "netcat" on the server and client sides,
measuring bandwidth with both "iptraf" and "gkrellm", and transfer rates
are the ones configured.

The configuration used in my test follows. Traffic was generated on the
client side via "cat /large/archive | nc remote_ip 8000", and received
on the server side with "nc -l -p 8000 > /dev/null". Hope it helps.

#!/bin/bash
DOWNLINK00
UPLINK00
DEV=eth0
if [ "$1" = "status" ]
then
tc -s qdisc ls dev $DEV
tc -s class ls dev $DEV
exit
fi

# clean existing down- and uplink qdiscs, hide
errors
tc qdisc del dev $DEV root    2> /dev/null >
/dev/null
tc qdisc del dev $DEV ingress 2> /dev/null >
/dev/null

if [ "$1" = "stop" ]
then
exit
fi

###### uplink

# install root HTB, point default traffic to 1:20:
tc qdisc add dev $DEV root handle 1: htb default 20
# shape everything at $UPLINK speed - this prevents
huge queues in your
# DSL modem which destroy latency:
tc class add dev $DEV parent 1: classid 1:1 htb rate
${UPLINK}kbit burst 6k
# high prio class 1:10:
tc class add dev $DEV parent 1:1 classid 1:10 htb
rate ${UPLINK}kbit \
   burst 6k prio 1
# bulk & default class 1:20 - gets slightly less
traffic, 
# and a lower priority:
tc class add dev $DEV parent 1:1 classid 1:20 htb
rate $[9*$UPLINK/10]kbit \
   burst 6k prio 2
tc class add dev $DEV parent 1:1 classid 1:30 htb
rate $[8*$UPLINK/10]kbit \
   burst 6k prio 2
# all get Stochastic Fairness:
tc qdisc add dev $DEV parent 1:10 handle 10: sfq
perturb 10
tc qdisc add dev $DEV parent 1:20 handle 20: sfq
perturb 10
tc qdisc add dev $DEV parent 1:30 handle 30: sfq
perturb 10
# TOS Minimum Delay (ssh, NOT scp) in 1:10:
tc filter add dev $DEV parent 1:0 protocol ip prio
10 u32 \
      match ip tos 0x10 0xff  flowid 1:10
# ICMP (ip protocol 1) in the interactive class 1:10
so we 
# can do measurements & impress our friends:
tc filter add dev $DEV parent 1:0 protocol ip prio
10 u32 \
        match ip protocol 1 0xff flowid 1:10
# To speed up downloads while an upload is going on,
put ACK packets in
# the interactive class:
tc filter add dev $DEV parent 1: protocol ip prio 10
u32 \
   match ip protocol 6 0xff \
   match u8 0x05 0x0f at 0 \
   match u16 0x0000 0xffc0 at 2 \
   match u8 0x10 0xff at 33 \
   flowid 1:10
# rest is 'non-interactive' ie 'bulk' and ends up in
1:20

tc filter add dev $DEV parent 1: protocol ip prio 18

u32 \
   match ip dst 0.0.0.0/0 flowid 1:20

########## downlink #############

# slow downloads down to somewhat less than the real
speed  to prevent 
# queuing at our ISP. Tune to see how high you can
set it.
# ISPs tend to have *huge* queues to make sure big
downloads are fast
#
# attach ingress policer:

tc qdisc add dev $DEV handle ffff: ingress

Quoted 13 lines
# filter *everything* to it (0.0.0.0/0), drop
everything that's
# coming in too fast:

tc filter add dev $DEV parent ffff: protocol ip prio

50 u32 match ip src \
   0.0.0.0/0 police rate ${DOWNLINK}kbit burst 10k

drop flowid :1

Signature · 3 lines
Jose Luis Domingo Lopez
Linux Registered User #189436     Debian Linux Sid
(Linux 2.5.68)

__________________________________
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.
http://search.yahoo.com
_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/