Congestion Control

1
Congestion Control
2
Congestion Control

• Too many packets in part of the network
• Performance Degrades Congestion
• Network Layer provides congestion control to
  ensure timely delivery of packets from source to
  destination

3
Congestion Control

• Causes of congestion
• Types of congestion control schemes
• Solving congestion

4
Causes of Congestion

• 1. Exhaustion of buffer space
• 2. Deadlock
• Congestion can be long-lived or transient

5
Exhaustion of Buffer Space

• Routers maintain packet queues (or buffers)
• Buffers fill up if
• Routers are too slow, OR
• Combined input traffic rate exceeds the outgoing
  traffic rate
• Insufficient buffer space leads to congestion

6
Exhaustion ofBuffer Space (contd)
50 Mbps
Router
50 Mbps
50 Mbps
50 Mbps
Buffer
7
Deadlock (or lockup)

• The first router cannot proceed until the second
  router does something, and the second router
  cannot proceed until the first router does
  something
• Both routers come to a complete halt and stay
  that way forever

8
Types of Deadlock

• 1. Store and Forward Lockup
• Direct Store and Forward Lockup
• Indirect Store and Forward Lockup
• 2. Reassembly Lockup

9
Direct Store and Forward Lockup

• Simplest lockup between two routers
• Example
• Suppose router A has five buffers, all of which
  are queued for output to router B
• Similarly, router B has five buffers, all of
  which are queued for output to router A
• If there is flow control on the link between
  routers A and B, then neither router can accept
  any incoming packets from the other. They are
  both stuck.

10
Direct Store and Forward Lockup (contd)
Router A
To/from other routers
To/from other routers
Router B
To/from other routers
To/from other routers
11
Indirect Store and Forward Lockup

• The same thing can happen on a larger scale

12
Cycles in a Dependency Graph

• If we view the network as a graph with nodes and
  edges

A deadlock occurs when There is a cycle of
dependencies in the graph
13
Reassembly Lockup

• In some network layer implementations, the
  sending router must split messages into multiple
  network layer packets
• Receiving routers reassemble split up packets
  into a single packet
• If the receiving routers buffer fills up with
  incomplete packets, it cannot reassemble any more
  packets

14
Reassembly Lockup (contd)
Buffer size 16 packet segments
s0, p2
s1, p1
s2, p1
s3, p1
s0, p3
s1, p2
s2, p2
s3, p3
s0, p4
s1, p4
s2, p3
s3, p4
s0, p5
s1, p5
s2, p5
s3, p5
Input 0
Input 1
Input 2
Input 3
ssource, ppart
15
Types of Congestion Control

• Preventive
• The hosts and routers attempt to prevent
  congestion before it can occur
• Reactive
• The hosts and routers respond to congestion after
  it occurs and then attempt to stop it

16
Preventive and Reactive Control

• Preventive Techniques
• Resource reservation
• Leaky/Token bucket
• Isarithmic control
• Fair Queuing
• Reactive Techniques
• Load shedding
• Choke packets

17
Solving Congestion

• Host based
• Isarithmic Control
• Leaky Bucket
• Token Bucket
• Router Based
• Load Shedding
• Choke Packets
• Resource Reservation
• Fair Queuing

18
Load Shedding

• When a router becomes inundated with packets, it
  simply drops some
• Load Shedding

19
Intelligent Load Shedding

• Discarding packets does not need to be done
  randomly
• Router should take other information into account
• Possibilities
• Total packet dropping
• Priority discarding
• Age biased discarding

20
Total Packet Dropping

• When the buffer fills and a packet segment is
  dropped, drop all the rest of the segments from
  that packet, since they will be useless anyway
• Only works with routers that segment and
  reassemble packets

21
Priority Discarding

• Sources specify the priority of their packets
• When a packet is discarded, the router chooses a
  low priority packet
• Requires hosts to participate by labeling their
  packets with priority levels

22
Age Biased Discarding

• When the router has to discard a packet, it
  chooses the oldest one in its buffer
• This works well for multimedia traffic which
  requires short delays
• This may not work so well for data traffic, since
  more packets will need to be retransmitted

23
Random Early Detection

• (RED)
• TCP detects packet loss and slows the sending
  rate accordingly
• When the router queues start to fill, randomly
  drop some packets

24
Choke Packets

• Each router monitors the utilization of each of
  its output lines
• Associated with each line is a variable u, which
  reflects the utilization of that line
• Whenever u moves above a given threshold value,
  the output line enters a warning state
• Each newly arriving packet checks if its output
  line is in the warning state
• If so, the router sends a choke packet back to
  the source

25
Choke Packets (contd)

• The data packet is tagged (by setting a bit in
  its header) so that it will not generate any more
  choke packets at downstream routers
• When the source host receives the choke packet,
  it is required to reduce its traffic generation
  rate to the specified destination by X
• Since other packets aimed at the same destination
  are probably already on their way to the
  congested location, the source host should ignore
  choke packets for that destination for a fixed
  time interval. After that, it resumes its
  response to choke packets.

26
Choke Packets Example
Congestion (u gtThr)
Host
router
router
Tagged Packets (to prevent further choke packet
generation at later routers)
Choke Packet
27
Isarithmic Control

• Each host is initially allocated a pool of
  permits
• Data packets carry permits between hosts
• A host cant send a packet unless at has at least
  one permit
• If no permit is available, the packet waits in
  the host until a permit becomes available.

28
Isarithmic ControlExample
3 Permits
0 Permits
A
B
Time 1
Data
2 Permits
0 Permits
A
B
Time 2
2 Permits
1 Permit
A
B
Time 3
29
Isarithmic Control (contd)

• Isarithmic control is implemented at the hosts,
  not the routers
• Isarithmic control guarantees that the number of
  packets in the network will never exceed the
  number of permits initially present

30
Isarithmic Control Problem

• If a host doesnt send any packet, i.e. it just
  holds onto its permits, then the network will not
  be utilized fully
• Solution Place a limit on the number of permits
  a host may keep. If the host has more than its
  allowed number of permits, it must transmit the
  excess permits to other hosts by piggybacking
  them onto other data packets or by using special
  permit-transport packets

31
Isarithmic ControlExample
Max-Permits 4
Data
1 Permit
3 Permits
A
B
Time 1
1 Permit
4 Permits
A
B
Time 2
Permit-transport Packet
1 Permit
3 Permits
A
B
Time 3
32
Resource Reservation

• Primarily for connection-oriented networks
• During connection setup
• Request resources (e.g., buffer space, connection
  bandwidth) from the network
• If the network has enough available resources to
  support the new connection, the connection will
  be established
• Otherwise, the connection will be rejected

33
Resource Reservation Example
Router
Router
Src
Dest
4 Mbps available
6 Mbps available
10 Mbps available
Case 1 Source attempts to connect to
destination, and attempts to
reserve 4 Mbps for the connection Result
Connection accepted. There is enough bandwidth
available. Available link
bandwidths updated. Case 2 Source attempts to
connect to destination, and attempts to
reserve 5 Mbps for the connection Result
Failure. There is not enough bandwidth available
on one of the links.
34
Resource Reservation (contd)

• Once a connection is accepted, the host must use
  only the amount of resources reserved. It may
  not use more than that.
• What if the host is malicious and attempts to use
  more network resources than it reserved?

35
Leaky Bucket

• Used in conjunction with resource reservation to
  police the hosts reservation
• At the host-network interface, allow packets into
  the network at a constant rate
• Packets may be generated in a bursty manner, but
  after they pass through the leaky bucket, they
  enter the network evenly spaced

36
Leaky Bucket Analogy
Packets from host
Leaky Bucket
Network
37
Leaky Bucket (contd)

• The leaky bucket is a traffic shaper It
  changes the characteristics of packet stream
• Traffic shaping makes more manageable and more
  predictable
• Usually the network tells the leaky bucket the
  rate at which it may send packets when the
  connection begins

38
Leaky Bucket Doesnt allow bursty transmissions

• In some cases, we may want to allow short bursts
  of packets to enter the network without smoothing
  them out
• For this purpose we use a token bucket, which is
  a modified leaky bucket

39
Token Bucket

• The bucket holds tokens instead of packets
• Tokens are generated and placed into the token
  bucket at a constant rate
• When a packet arrives at the token bucket, it is
  transmitted if there is a token available.
  Otherwise it is buffered until a token becomes
  available.
• The token bucket has a fixed size, so when it
  becomes full, subsequently generated tokens are
  discarded

40
Token Bucket
Packets from host
Token Generator (Generates a token once every T
seconds)
Network
41
Token Bucket vs. Leaky Bucket
Case 1 Short burst arrivals
Arrival time at bucket
6
5
4
3
2
1
0
Departure time from a leaky bucket Leaky bucket
rate 1 packet / 2 time units Leaky bucket size
4 packets
6
5
4
3
2
1
0
Departure time from a token bucket Token bucket
rate 1 tokens / 2 time units Token bucket size
2 tokens
6
5
4
3
2
1
0
42
Token Bucket vs. Leaky Bucket
Case 2 Large burst arrivals
Arrival time at bucket
6
5
4
3
2
1
0
Departure time from a leaky bucket Leaky bucket
rate 1 packet / 2 time units Leaky bucket size
2 packets
6
5
4
3
2
1
0
Departure time from a token bucket Token bucket
rate 1 token / 2 time units Token bucket size
2 tokens
6
5
4
3
2
1
0
43
Router based congestion management

• Isarithmic, Token bucket and leaky bucket manage
  traffic from the end host.
• But what if we do not trust the incomming traffic
  to behave?
• Must manage within the router

44
Fair queuing

• If one source is sending too many packets, can we
  allow other sources to continue and just drop the
  bad source?
• First cut round-robin

45
Round-Robin
46

• Advantages protection among flows
• Misbehaving flows will not affect the performance
  of well-behaving flows
• Misbehaving flow a flow that does not implement
  congestion control
• Disadvantages
• More complex must maintain a queue per flow per
  output instead of a single queue per output
• Biased toward large packets a flow receives
  service proportional to the number of packets

47
Idealized flow model

• For N sources, we would like to give each host or
  input source 1/N of the link bandwidth
• Image we could squeeze factions of a bits on the
  link at once (fluid model)
• Solution would interleave the bits on the link at
  the same instant

48
Fluid model vs. Packet Model
49
Approximate bit-bit RR

• Virtual time is incremented each time a bit is
  transmitted for all flows
• If we have 3 active flows, and transmit 3 bits,
  we increment virtual time by 1.
• At each packet arrival, compute time packet would
  have exited the router during virtual time.
• This is the packet finish time

50
Approximate bit-RR

• Algorithm selects the queue with the packet that
  has the lowest finish time
• This is fair queuing