P13ITE05

High Speed Networks

UNIT - IV

Dr.A.Kathirvel
Professor & Head/IT - VCEW
 UNIT - IV
 Integrated services architecture
 Approach, components and services
 Queuing discipline: FQ, PS, BRFQ, WFQ
 Random early detection
 Differentiated services
 Introduction
 New additions to Internet increasing traffic
 High volume client/server application
 Web
 Graphics

 Real time voice and video

 Need to manage traffic and control congestion
 IEFT standards
 Integrated services
 Collective service to set of traffic demands in domain

 Limit demand & reserve resources

 Differentiated services
 Classify traffic in groups

 Different group traffic handled differently

3
Integrated Services Architecture (ISA)‫‏‬

 IPv4 header fields for precedence and type of service

usually ignored
 ATM only network designed to support TCP, UDP and
real-time traffic
 May need new installation
 Need to support Quality of Service (QoS) within TCP/IP
 Add functionality to routers
 Means of requesting QoS

4
 Internet Traffic – Elastic
 Can adjust to changes in delay and throughput
 E.g. common TCP and UDP application
 E-Mail – insensitive to delay changes
 FTP – User expect delay proportional to file size
 Sensitive to changes in throughput

 SNMP – delay not a problem, except when caused by congestion

 Web (HTTP), TELNET – sensitive to delay
 Not per packet delay – total elapsed time
 E.g. web page loading time
 For small items, delay across internet dominates
 For large items it is throughput over connection
 Need some QoS control to match to demand

5
 Internet Traffic – Inelastic
 Does not easily adapt to changes in delay and throughput
 Real time traffic
 Throughput
 Minimum may be required
 Delay
 E.g. stock trading
 Jitter - Delay variation
 More jitter requires a bigger buffer
 E.g. teleconferencing requires reasonable upper bound
 Packet loss

6
 Inelastic Traffic Problems
 Difficult to meet requirements on network with
variable queuing delays and congestion
 Need preferential treatment
 Applications need to state requirements
 Ahead of time (preferably) or on the fly
 Using fields in IP header
 Resource reservation protocol
 Must still support elastic traffic
 Deny service requests that leave too few resources
to handle elastic traffic demands

7
 ISA Approach
 Provision of QoS over IP
 Sharing available capacity when congested
 Router mechanisms
 Routing Algorithms
 Select to minimize delay

 Packet discard
 Causes TCP sender to back off and reduce load

 Enahnced by ISA

8
 Flow
 IP packet can be associated with a flow
 Distinguishable stream of related IP packets
 From single user activity
 Requiring same QoS
 E.g. one transport connection or one video stream
 Unidirectional
 Can be more than one recipient
 Multicast

 Membership of flow identified by source and destination IP

address, port numbers, protocol type
 IPv6 header flow identifier can be used but isnot
necessarily equivalent to ISA flow

9
 ISA Functions
 Admission control
 For QoS, reservation required for new flow
 RSVP used
 Routing algorithm
 Base decision on QoS parameters
 Queuing discipline
 Take account of different flow requirements
 Discard policy
 Manage congestion
 Meet QoS

10
 ISA Implementation in Router
 Background
Functions

 Forwarding
functions

11
ISA Components – Background Functions
 Reservation Protocol
 RSVP
 Admission control
 Management agent
 Can use agent to modify traffic control database and
direct admission control
 Routing protocol

12
 ISA Components – Forwarding
 Classifier and route selection
 Incoming packets mapped to classes
 Single flow or set of flows with same QoS

E.g. all video flows

 Based on IP header fields

 Determines next hop

 Packet scheduler
 Manages one or more queues for each output
 Order queued packets sent
 Based on class, traffic control database, current

and past activity on outgoing port

 Policing
13
 ISA Services

 Traffic specification (TSpec) defined as service for flow

 On two levels
 General categories of service
 Guaranteed

 Controlled load

 Best effort (default)‫‏‬

 Particular flow within category

 TSpec is part of contract

14
 Token Bucket

 Many traffic sources can be defined by token

bucket scheme
 Provides concise description of load imposed by
flow
 Easy to determine resource requirements
 Provides input parameters to policing function

15
Token Bucket Diagram

16
 ISA Services - Guaranteed Service
 Assured capacity level or data rate
 Specific upper bound on queuing delay through
network
 Must be added to propagation delay or latency to
get total delay
 Set high to accommodate rare long queue delays
 No queuing losses
 I.e. no buffer overflow
 E.g. Real time play back of incoming signal can use
delay buffer for incoming signal but will not tolerate
packet loss

17
 ISA Services - Controlled Load
 Tightly approximates to best efforts under unloaded conditions
 No upper bound on queuing delay
 High percentage of packets do not experience delay over
minimum transit delay
 Propagation plus router processing with no queuing

delay
 Very high percentage delivered
 Almost no queuing loss
 Adaptive real time applications
 Receiver measures jitter and sets playback point
 Video can drop a frame or delay output slightly
 Voice can adjust silence periods

18
 Queuing Discipline
 Traditionally first in first out (FIFO) or first come first
served (FCFS) at each router port
 No special treatment to high priority packets (flows)‫‏‬
 Small packets held up by large packets ahead of them
in queue
 Larger average delay for smaller packets
 Flows of larger packets get better service
 Greedy TCP connection can crowd out altruistic
connections
 If one connection does not back off, others may
back off more

19
 Fair Queuing (FQ)‫‏‬
 Multiple queues for each port
 One for each source or flow
 Queues services round robin
 Each busy queue (flow) gets exactly one packet per
cycle
 Load balancing among flows
 No advantage to being greedy
 Your queue gets longer, increasing your delay

 Short packets penalized as each queue sends one

packet per cycle

20
FIFO and FQ

21
 Processor Sharing

 Multiple queues as in FQ
 Send one bit from each queue per round
 Longer packets no longer get an advantage
 Can work out virtual (number of cycles) start and
finish time for a given packet
 However, we wish to send packets, not bits

22
 Bit-Round Fair Queuing (BRFQ)‫‏‬

 Compute virtual start and finish time as before

 When a packet finished, the next packet sent is the
one with the earliest virtual finish time
 Good approximation to performance of PS
 Throughput and delay converge as time increases

23
Examples
of PS and
BRFQ

24
Comparison of FIFO, FQ and BRFQ

25
Generalized Processor Sharing (GPS)‫‏‬

 BRFQ can not provide different capacities to different

flows
 Enhancement called Weighted fair queue (WFQ)‫‏‬
 From PS, allocate weighting to each flow that determines
how many bots are sent during each round
 If weighted 5, then 5 bits are sent per round
 Gives means of responding to different service requests
 Guarantees that delays do not exceed bounds

26
 Weighted Fair Queue

 Emulates bit by bit GPS

 Same strategy as BRFQ

27
FIFO And WFQ

Chapter 17 Integrated and Differentiated 28

Services
 Proactive Packet Discard

 Congestion management by proactive packet discard

 Before buffer full
 Used on single FIFO queue or multiple queues for
elastic traffic
 E.g. Random Early Detection (RED)‫‏‬

29
Random Early Detection(RED) Motivation
 Surges fill buffers and cause discards
 On TCP this is a signal to enter slow start phase, reducing load
 Lost packets need to be resent
 Adds to load and delay

 Global synchronization
 Traffic burst fills queues so packets lost

 Many TCP connections enter slow start

 Traffic drops so network under utilized

 Connections leave slow start at same time causing burst

 Bigger buffers do not help

 Try to anticipate onset of congestion and tell one connection to slow
down

30
 RED Design Goals

 Congestion avoidance
 Global synchronization avoidance
 Current systems inform connections to back off
implicitly by dropping packets
 Avoidance of bias to bursty traffic
 Discard arriving packets will do this
 Bound on average queue length
 Hence control on average delay

31
 RED Algorithm – Overview
Calculate average queue size avg
if avg < THmin
queue packet
else if THmin  avg  Thmax
calculate probability Pa
with probability Pa
discard packet
else with probability 1-Pa
queue packet
else if avg  THmax
discard packet

32
RED Buffer

33
RED Algorithm Detail

34
 Differentiated Services (DS)‫‏‬
 ISA and RSVP complex to deploy
 May not scale well for large volumes of traffic
 Amount of control signals
 Maintenance of state information at routers
 DS architecture designed to provide simple, easy
to implement, low overhead tool
 Support range of network services
 Differentiated on basis of performance

35
 Characteristics of DS
 Use IPv4 header Type of Service or IPv6 Traffic Class field
 No change to IP
 Service level agreement (SLA) established between provider
(internet domain) and customer prior to use of DS
 DS mechanisms not needed in applications
 Build in aggregation
 All traffic with same DS field treated same
 E.g. multiple voice connections

 DS implemented in individual routers by queuing and

forwarding based on DS field
 State information on flows not saved by routers

36
DS Terminology

Chapter 17 Integrated and Differentiated 37

Services
 Services
 Provided within DS domain
 Contiguous portion of Internet over which consistent set of
DS policies administered
 Typically under control of one administrative entity
 Defined in SLA
 Customer may be user organization or other DS domain
 Packet class marked in DS field
 Service provider configures forwarding policies routers
 Ongoing measure of performance provided for each class
 DS domain expected to provide agreed service internally
 If destination in another domain, DS domain attempts to forward
packets through other domains
 Appropriate service level requested from each domain

38
 SLA Parameters

 Detailed service performance parameters

 Throughput, drop probability, latency
 Constraints on ingress and egress points
 Indicate scope of service
 Traffic profiles to be adhered to
 Token bucket
 Disposition of traffic in excess of profile

39
 Example Services
 Qualitative
 A: Low latency
 B: Low loss
 Quantitative
 C: 90% in-profile traffic delivered with no more than 50ms
latency
 D: 95% in-profile traffic delivered
 Mixed
 E: Twice bandwidth of F
 F: Traffic with drop precedence X has higher delivery
probability than that with drop precedence Y

40
DS Field v IPv4 Type of Service

41
 DS Field Detail
 Leftmost 6 bits are DS codepoint
 64 different classes available
 3 pools
 xxxxx0 : reserved for standards

 000000 : default packet class

 xxx000 : reserved for backwards compatibility with
IPv4 TOS
 xxxx11 : reserved for experimental or local use

 xxxx01 : reserved for experimental or local use but may

be allocated for future standards if needed

 Rightmost 2 bits unused

42
Configuration Diagram

43
 Configuration – Interior Routers

 Domain consists of set of contiguous routers

 Interpretation of DS codepoints within domain is consistent
 Interior nodes (routers) have simple mechanisms to handle
packets based on codepoints
 Queuing gives preferential treatment depending on
codepoint
 Per Hop behaviour (PHB)‫‏‬

 Must be available to all routers

 Typically the only part implemented in interior routers

 Packet dropping rule dictated which to drop when buffer

saturated

44
 Configuration – Boundary Routers
 Include PHB rules
 Also traffic conditioning to provide desired service
 Classifier
 Separate packets into classes

 Meter
 Measure traffic for conformance to profile

 Marker
 Policing by remarking codepoints if required

 Shaper
 Dropper

45
DS Traffic Conditioner

46
Per Hop Behaviour – Expedited forwarding
 Premium service
 Low loss, delay, jitter; assured bandwidth end-to-end
service through domains
 Looks like point to point or leased line
 Difficult to achieve
 Configure nodes so traffic aggregate has well defined
minimum departure rate
 EF PHB

 Condition aggregate so arrival rate at any node is

always less that minimum departure rate
 Boundary conditioners

47
Per Hop Behaviour – Explicit Allocation
 Superior to best efforts
 Does not require reservation of resources
 Does not require detailed discrimination among flows
 Users offered choice of number of classes
 Monitored at boundary node
 In or out depending on matching profile or not
 Inside network all traffic treated as single pool of
packets, distinguished only as in or out
 Drop out packets before in packets if necessary
 Different levels of service because different number of
in packets for each user

48
 PHB - Assured Forwarding

 Four classes defined

 Select one or more to meet requirements
 Within class, packets marked by customer or provider
with one of three drop precedence values
 Used to determine importance when dropping
packets as result of congestion

49
Codepoints for AF PHB

50
Questions

51