Friday, April 16, 2010

tc: Linux HTTP Outgoing Traffic Shaping (Port 80 Traffic Shaping)

I've 10Mbps server port dedicated to our small business server. The server also act as a backup DNS server and I'd like to slow down outbound traffic on port 80.

How do I limit bandwidth allocation to http service 5Mbps (burst to 8Mbps) at peak times so that DNS and other service will not go down due to heavy activity under Linux operating systems?


You need use the tc command which can slow down traffic for given port and services on servers and it is called traffic shaping:
  1. When traffic is shaped, its rate of transmission is under control, in other words you apply some sort of bandwidth allocation for each port or or so called Linux services. Shaping occurs on egress.
  2. You can only apply traffic shaping to outgoing or forwarding traffic i.e. you do not have any control for incoming traffic to server. However, tc can do policing controls for arriving traffic. Policing thus occurs on ingress. This FAQ only deals with traffic shaping.

Token Bucket (TB)

A token bucket is nothing but a common algorithm used to control the amount of data that is injected into a network, allowing for bursts of data to be sent.

It is used for network traffic shaping or rate limiting. With token bucket you can define the maximum rate of traffic allowed on an interface at a given moment in time.

tokens/sec
                                   |   |
                                   |  | Bucket to
                                   |  | to hold b tokens
                             +======+=====+
                                          |
                                          |
        |                                \|/
Packets |      +============+
stream  | ---> | token wait | --->  Remove token  --->  eth0
        |      +============+
  1. The TB filter puts tokens into the bucket at a certain rate.
  2. Each token is permission for the source to send a specific number of bits into the network.
  3. Bucket can hold b tokens as per shaping rules.
  4. Kernel can send packet if you've a token else traffic need to wait.

How Do I Use tc command?

WARNING! These examples requires good understanding of TCP/IP and other networking concepts. All new user should try out examples in test environment.

tc command is by default installed on my Linux distributions. To list existing rules, enter:


# tc -s qdisc ls dev eth0


Sample outputs:
qdisc pfifo_fast 0: root bands 3 priomap  1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
 Sent 2732108 bytes 10732 pkt (dropped 0, overlimits 0 requeues 0)
 rate 0bit 0pps backlog 0b 0p requeues 0

Your First Traffic Shaping Rule

First, send ping request to cyberciti.biz from your Local Linux workstation and note down ping time, enter:


# ping cyberciti.biz


Sample outputs:
PING cyberciti.biz (74.86.48.99) 56(84) bytes of data.
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=1 ttl=47 time=304 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=2 ttl=47 time=304 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=3 ttl=47 time=304 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=4 ttl=47 time=304 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=5 ttl=47 time=304 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=6 ttl=47 time=304 ms

Type the following tc command to slow down traffic by 200 ms:


# tc qdisc add dev eth0 root netem delay 200ms


Now, send ping requests again:


# ping cyberciti.biz


Sample outputs:
PING cyberciti.biz (74.86.48.99) 56(84) bytes of data.
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=1 ttl=47 time=505 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=2 ttl=47 time=505 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=3 ttl=47 time=505 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=4 ttl=47 time=505 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=5 ttl=47 time=505 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=6 ttl=47 time=505 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=7 ttl=47 time=505 ms
64 bytes from txvip1.simplyguide.org (74.86.48.99): icmp_seq=8 ttl=47 time=505 ms
^C
--- cyberciti.biz ping statistics ---
8 packets transmitted, 8 received, 0% packet loss, time 7006ms
rtt min/avg/max/mdev = 504.464/505.303/506.308/0.949 ms

To list current rules, enter:


# tc -s qdisc ls dev eth0


Sample outputs:
qdisc netem 8001: root limit 1000 delay 200.0ms
 Sent 175545 bytes 540 pkt (dropped 0, overlimits 0 requeues 0)
 rate 0bit 0pps backlog 0b 0p requeues 0 

To delete all rules, enter:


# tc qdisc del dev eth0 root
# tc -s qdisc ls dev eth0



TBF Example
To attach a TBF with a sustained maximum rate of 1mbit/s, a peakrate of 2.0mbit/s, a 10kilobyte buffer, with a pre-bucket queue size limit calculated so the TBF causes at most 70ms of latency, with perfect peakrate behavior, enter:


# tc qdisc add dev eth0 root tbf rate 1mbit burst 10kb latency 70ms peakrate 2mbit minburst 1540


HTB - Hierarchy Token Bucket
To control the use of the outbound bandwidth on a given link use HTB:
  1. rate - You can set the allowed bandwidth.
  2. ceil - You can set burst bandwidth allowed when buckets are present.
  3. prio - You can set priority for additional bandwidth. So classes with lower prios are offered the bandwidth first. For example, you can give lower prio for DNS traffic and higher for HTTP downloads.
  4. iptables and tc: You need to use iptables and tc as follows to control outbound HTTP traffic.

Example: HTTP Outbound Traffic Shaping

First , delete existing rules for eth1:


# /sbin/tc qdisc del dev eth1 root


Turn on queuing discipline, enter:


# /sbin/tc qdisc add dev eth1 root handle 1:0 htb default 10


Define a class with limitations i.e. set the allowed bandwidth to 512 Kilobytes and burst bandwidth to 640 Kilobytes for port 80:


# /sbin/tc class add dev eth1 parent 1:0 classid 1:10 htb rate 512kbps ceil 640kbps prio 0


Please note that port 80 is NOT defined anywhere in above class. You will use iptables mangle rule as follows:


# /sbin/iptables -A OUTPUT -t mangle -p tcp --sport 80 -j MARK --set-mark 10


To save your iptables rules, enter (RHEL specific command):


# /sbin/service iptables save


Finally, assign it to appropriate qdisc:


# tc filter add dev eth1 parent 1:0 prio 0 protocol ip handle 10 fw flowid 1:10


Here is another example for port 80 and 22:


/sbin/tc qdisc add dev eth0 root handle 1: htb
/sbin/tc class add dev eth0 parent 1: classid 1:1 htb rate 1024kbps
/sbin/tc class add dev eth0 parent 1:1 classid 1:5 htb rate 512kbps ceil 640kbps prio 1
/sbin/tc class add dev eth0 parent 1:1 classid 1:6 htb rate 100kbps ceil 160kbps prio 0
/sbin/tc filter add dev eth0 parent 1:0 prio 1 protocol ip handle 5 fw flowid 1:5
/sbin/tc filter add dev eth0 parent 1:0 prio 0 protocol ip handle 6 fw flowid 1:6
/sbin/iptables -A OUTPUT -t mangle -p tcp --sport 80 -j MARK --set-mark 5
/sbin/iptables -A OUTPUT -t mangle -p tcp --sport 22 -j MARK --set-mark 6



How Do I Monitor And Test Speed On Sever?
Use the following tools


# /sbin/tc -s -d class show dev eth0
# /sbin/iptables -t mangle -n -v -L
# iptraf
# watch /sbin/tc -s -d class show dev eth0



To test download speed use lftp or wget command line tools.

1 comment: