Data and Computer

Communications
Chapter 13 – Congestion in Data
Networks

Eighth Edition
by William Stallings

Lecture slides by Lawrie Brown

 Congestion in Data Networks

At St. Paul's a great throng crammed the platform. She

saw a sea of faces, each stamped with a kind of
purposeful, hungry urgency, a determination to get into
this train. As before, when she was on the Northern
Line, she thought there must be some rule, some
operating law, that would stop more than a limited,
controlled number getting in. Authority would appear
and stop it.
—King Solomon's Carpet, Barbara Vine (Ruth Rendell)
 What Is Congestion?
 congestion occurs when the no of packets being
transmitted through the network approaches the
packet handling capacity of the network
 congestion control aims to keep no of packets
below a level at which performance falls off
dramatically
 a data network is a network of queues
 generally 80% utilization is critical
 finite queues mean data may be lost
Queues at a Node
Interaction of Queues
 Ideal
Network
Utilization
Effects of
Congestion
-
No Control
 Mechanisms for
Congestion Control
 Backpressure
 if node becomes congested it can slow down or
halt flow of packets from other nodes
 cf. backpressure in blocked fluid pipe
 may mean that other nodes have to apply control on
incoming packet rates
 propagates back to source
 can restrict to high traffic logical connections
 used in connection oriented nets that allow hop
by hop congestion control (eg. X.25)
 not used in ATM nor frame relay
 only recently developed for IP
 Choke Packet
 a control packet
 generated at congested node
 sent to source node
 eg. ICMP source quench
• from router or destination
• source cuts back until no more source quench message
• sent for every discarded packet, or anticipated
 is a rather crude mechanism
Implicit Congestion Signaling
 transmission delay increases with congestion
 hence a packet may be discarded
 source detects this implicit congestion indication
 useful on connectionless (datagram) networks
 eg. IP based
• (TCP includes congestion and flow control - see chapter 17)
 used in frame relay LAPF
Explicit Congestion Signaling
 network alerts end systems of increasing
congestion
 end systems take steps to reduce offered load
 Backwards
 congestion avoidance notification in opposite direction
to packet required
 Forwards
 congestion avoidance notification in same direction as
packet required
Explicit Signaling Categories
 Binary
 a bit set in a packet indicates congestion
 Credit based
 indicates how many packets source may send
 common for end to end flow control
 Rate based
 supply explicit data rate limit
 nodes along path may request rate reduction
 eg. ATM
 Traffic Management
 fairness
 provide equal treatment of various flows
 quality of service
 different treatment for different connections
 reservations
 traffic contract between user and network
 carry best-effort or discard excess traffic
Congestion Control in Packet
Switched Networks
 send control packet to some or all source
nodes
 requires additional traffic during congestion
 rely on routing information
 may react too quickly
 end to end probe packets
 adds to overhead
 add congestion info to packets in transit
 either backwards or forwards
 Frame Relay
Congestion Control
 minimize discards
 maintain agreed QoS
 minimize probability of one end user monopoly
 simple to implement
 create minimal additional traffic
 distribute resources fairly
 limit spread of congestion
 operate effectively regardless of traffic flow
 minimum impact on other systems
 minimize variance in QoS
 FR Control Techniques
 difficult for frame-relay
 joint network & end-system responsibility
 techniques:
 discard strategy
 congestion avoidance
 explicit signaling
 congestion recovery
 implicit signaling mechanism
 Traffic Rate Management
 must discard frames to cope with congestion
 arbitrarily, no regard for source
 no reward for restraint so end systems transmit as
fast as possible
 Committed information rate (CIR)
• data in excess of this liable to discard
• not guaranteed in extreme congestion situations
• aggregate CIR should not exceed physical data rate
 Committed burst size
 Excess burst size
Operation of CIR
Relationship

Among
Congestion
Parameters
Congestion Avoidance using
Explicit Signaling
 network alerts end systems of growing
congestion using
 backward explicit congestion notification
 forward explicit congestion notification
 frame handler monitors its queues
 may notify some or all logical connections
 user response reduce rate
 ATM Traffic Management
 high speed, small cell size, limited overhead bits
 still evolving
 reasons existing tools are inadequate for ATM
 majority of traffic not amenable to flow control
 feedback slow due to reduced transmission time
compared with propagation delay
 wide range of application demands
 different traffic patterns
 different network services
 high speed switching and transmission increases
volatility
 Latency/Speed Effects
 consider ATM at 150Mbps
 takes ~2.8x10-6 seconds to insert single cell
 time to traverse network depends on propagation
delay and switching delay
 assume propagation at two-thirds speed of light
 if source and destination on opposite sides of USA,
propagation time ~ 48x10-3 seconds
 given implicit congestion control, by the time dropped
cell notification has reached source, 7.2x106 bits have
been transmitted
 this is not a good strategy for ATM
 Cell Delay Variation
 for ATM voice/video, data is a stream of cells
 delay across network must be short
 rate of delivery must be constant
 there will always be some variation in transit
 delay cell delivery to application so that constant
bit rate can be maintained to application
Timing of CBR Cells
 Network Contribution to
Cell Delay Variation
 in packet switched networks is due to
queuing delays and routing decision time
 in Frame relay networks is similar
 in ATM networks
 less than frame relay
 ATM protocol designed to minimize
processing overheads at switches
 ATM switches have very high throughput
 only noticeable delay is from congestion
 must not accept load that causes congestion
 Cell Delay Variation
At The UNI
 application produces data at fixed rate
 3 layers of ATM processing causes delay
 interleaving cells from different connections
 operation and maintenance cell interleaving
 if using synchronous digital hierarchy frames,
these are inserted at physical layer
 cannot predict these delays
Origins of Cell Delay Variation
 Traffic and Congestion
Control Framework
 ATM layer traffic and congestion control
should support QoS classes for all
foreseeable network services
 should not rely on AAL protocols that are
network specific, nor higher level
application specific protocols
 should minimize network and end to end
system complexity
 Timings Considered
 timing intervals considered:
 cell insertion time
 round trip propagation time
 connection duration
 long term
 traffic control strategy then must:
 determine whether a given new connection can be
accommodated
 agree performance parameters with subscriber
 now review various control techniques
Resource Management Using
Virtual Paths
 separate traffic flow according to service
characteristics on a virtual path
 user to user application
 user to network application
 network to network application
 QoS parameters concerned with are:
 cell loss ratio
 cell transfer delay
 cell delay variation
Configuration of
VCCs and VPCs
 Allocating VCCs within VPC
 all VCCs within VPC should experience
similar network performance
 options for allocation:
 aggregate peak demand
• set VPC capacity to total of all peak VCC rates
• will meet peak demands, but often underutilized
 statistical multiplexing
• set VPC capacity to more than average VCC rates
• will see greater variation but better utilization
 Connection Admission
Control
 first line of defense
 user specifies traffic characteristics for new
connection (VCC or VPC) by selecting a QoS
 network accepts connection only if it can meet
the demand
 traffic contract
 peak cell rate
 cell delay variation
 sustainable cell rate
 burst tolerance
 Usage Parameter Control
 UPC function monitors a connection to
ensure traffic obeys contract
 purpose is to protect network resources
from overload by one connection
 done on VCC and VPC
 peak cell rate and cell delay variation
 sustainable cell rate and burst tolerance
 UPC discards cells outside traffic contract
 Selective Call Discard
 when network at point beyond UPC
discards (CLP=1) cells
 aim to discard lower-priority cells when
congested to protect higher-priority cells
 note. can’t distinguish between cells originally
labeled lower priority, verses those tagged by
UPC function
 Traffic Shaping
 UPC provides a form of traffic policing
 can be desirable to also shape traffic
 smoothing out traffic flow
 reducing cell clumping
 token bucket
Token Bucket for
Traffic Shaping
 GFR Traffic Management
 guaranteed frame rate (GFR) as simple as UBR
from end system viewpoint
 places modest requirements on ATM network
 end system does no policing or shaping of traffic
 may transmit at line rate of ATM adaptor
 no guarantee of delivery
 so higher layer (eg. TCP) must do congestion control
 user can reserve capacity for each VC
 ensures application can send at min rate with no loss
 if no congestion, higher rates maybe used
 Frame Recognition
 GFR recognizes frames as well as cells
 when congested, network discards whole frame
rather than individual cells
 all cells of a frame have same CLP bit setting
 CLP=1 AAL5 frames lower priority (best effort)
 CLP=0 frames minimum guaranteed capacity
 GFR Contract Parameters
 Peak cell rate (PCR)
 Minimum cell rate (MCR)
 Maximum burst size (MBS)
 Maximum frame size (MFS)
 Cell delay variation tolerance (CDVT)
Components of GFR System
Supporting Rate Guarantees
 Tagging and Policing
 discriminates between frames that
conform to contract and those that don’t
 set CLP=1 on all cells in frame if not
 gives lower priority
 maybe done by network or source
 network may discard CLP=1 cells
 policing
 Buffer Management
 deals with treatment of buffered cells
 congestion indicated by high buffer
occupancy
 will discard tagged cells in preference to
untagged cells
 including ones already in buffer to make room
 may do per VC buffering for fairness
 cell discard based on queue-specific
thresholds
 Scheduling
 preferential treatment to untagged cells
 separate queues for each VC
 make per-VC scheduling decisions
 enables control of outgoing rate of VCs
 VCs get fair capacity allocation
 still meet contract
GFC Conformance Definition
 UPC function monitors each active VC
 to ensure traffic conforms to contract
 tag or discard nonconforming cells
 frame conforms if all cells conform
 a cell conforms if:
 rate of cells is within contract
 all cells in frame have same CLP
 frame satisfies MFS parameter
• check if either last cell in frame or cell count < MFS
 QoS Eligibility Test
 two stage filtering process
 a frame is tested for conformance to contract
 if not, may discard or tag
 set upper bound & penalize cells above upper bound
 do expect attempt to deliver tagged cells
 determine frames eligible for QoS guarantees
 under GFR contract for VC
 set lower bound on traffic
 frames in traffic flow below threshold are eligible
 GFR VC Frame Categories
 nonconforming frame
 cells of this frame will be tagged or discarded
 conforming but ineligible frames
 cells will receive a best-effort service
 conforming and eligible frames
 cells will receive a guarantee of delivery 

 form of cell rate algorithm is used

 Summary
 congestion effects
 congestion control
 traffic management
 frame relay congestion control
 ATM congestion control