Packet Scheduling

1
Packet Scheduling

• CS2520-TELCOM2321
• Wide Area Networks
• KyoungSoo Park
• University of Pittsburgh
• Slides borrowed from Dr. Cerroni

2
Todays Lecture

• Packet scheduling determine packet processing
  priority
• First come first served (FCFS/FIFO)
• Priority queuing
• (Weighted) fair queuing
• Max-min fair share algorithm
• Generalized processor sharing (GPS)
• Traffic policing vs. shaping
• Dropping/marking vs. spreading out
• Leaky bucket
• Token bucket

3
Multi-Service Network

• Single network infrastructure carrying traffic
  from different services (data, voice, video...)
  in an integrated way
• Different quality of service requirements to be
  met

4
Performance Bounds Delay Jitter

• Delay jitter bound is useful for audio/video
  playback, where receiver can compensate delay
  variations
• Delaying the first packet by the delay-jitter
  bound in an elasticity buffer
• Playing back packets at a constant rate from the
  buffer

reception at the client
Constant Bit Rate(CBR) streaming at the source
buffered data
Cumulative data
variable network delay
CBR playbackat the client
playbackbuffer delay
Time
5
Packet Scheduling

• Packets are queued at intermediate nodes
• Store and forward
• The server on each queue must
• Decide which packet to serve next
• Manage the queue of backlogged packets
• The server uses a scheduling discipline to
• Fairly share the resources
• Provide performance guarantees
• Trade-off between
• Traffic isolation
• Circuit switching guaranteed dedicated resources
• Resource sharing
• Packet switching complete resource sharing

6
Scheduling Policy Requirements

• A scheduler should be
• Scalable
• The number of operations to implement a
  discipline should be as independent as possible
  of the number of scheduled connections
• Fair
• It should allocate a fair share of the link
  capacity and output queue buffer to each
  connection
• Protective
• The misbehavior of one source should not affect
  the performance received by other sources

7
The Simplest Scheduler FCFS or FIFO

• First Come First Served or First In First Out
• Packets are served in the same order as they
  arrive
• Most commonly used scheduling discipline
• Example
• M/G/1 queue Poisson arrivals, generic service
  time
• Stability condition on server utilization
• Probability of idle server state

departure rate
arrival rate
averageservice time
server
queue
8
Limitations of FCFS

• No consideration for delay sensitivity
• Small packets must wait for long packets to be
  serviced
• Unfair
• Connections with large packets usually receive
  better services by getting higher percentage of
  server time
• Not protective
• Greedy connections are advantaged over friendlier
  connections
• An excessively active connection may cause other
  connections implementing congestion control to
  back off more than required
• Complete resource sharing without traffic
  isolation
• Isolation through separate queues for different
  traffic flows or classes sharing the same
  bandwidth

9
Priority Queuing

• K priority classes
• Class 1 has highest priority, then class 2, then
  class 3, ...
• Differentiate packet processing by its priority
• Low-priority packets are serviced only when there
  are no packets of higher priority waiting to be
  serviced
• FCFS within the same class
• Arrivals of high-priority packets do not stop
  current service of low-priority packets
  (non-preemptive)

10
Priority Queuing
...

• Flow with arrival rate and average
  service time
• server utilization due to
  flow packets
• Stability condition

11
Limitations of Priority Queuing

• Effective only when high priority traffic is a
  small part of the total traffic
• Efficient admission control and policing may be
  required
• Not protective
• A misbehaving source at a higher priority level
  can increase delay at lower priority levels
• Unfair
• Starvation of low-priority queues in case of
  heavy high-priority traffic
• More sophisticated scheduling disciplines are
  required
• Better traffic isolation
• Fairer resource sharing

12
Conservation Law

• Traffic isolation while sharing resources is
  limited by the conservation law
• The sum of the mean queuing delays received by
  the set of multiplexed connections, weighted by
  their share of the link's load, is independent of
  the scheduling discipline
• In other words, a scheduling discipline can
  reduce a particular connection's mean delay,
  compared with FCFS, only at the expense of
  another connection
• Valid if server is idle only when queues are
  empty
• Work-conserving schedulers

13
Max-Min Fair Share

• Resources should be shared in a fair way,
  especially when they are not sufficient to
  satisfy all requests
• Max-Min fair share
• Bandwidth allocation technique that tries to
  maximize the minimum share for non-satisfied
  flows
• Basic principles
• Resources are allocated in order of increasing
  demands
• Users with relatively small demands are satisfied
• No user gets a resource share larger than its
  demand
• Users with unsatisfied demands get an equal share
  of unused resources

14
Max-Min Fair Share

• Order demands d1, d2, ..., dN such that d1 d2
  ... dN
• Compute the initial fair share of the total
  capacity CFS C / N
• For each demand dk such that dk FS allocate the
  requested bandwidth
• Compute the remaining bandwidth R and the number
  of unsatisfied demands M
• Repeat steps 3 and 4 using the updated fair share
  FS R / M until no demand left is FS
• Allocate the current FS to remaining demands

15
Max-Min fair share Example

• Demands d1 23, d2 27, d3 35, d4 45, d5
  55
• Bandwidth C 155
• FS 155 / 5 31
• d1 and d2 satisfied, M 3
• Bandwidth left R 155 (23 27) 105
• FS 105 / 3 35
• d3 satisfied, M 2
• R 105 35 70
• FS 70 / 2 35
• d4 and d5 not satisfied ? they get 35

16
Max-Min Weighted Fair Share

• Max-Min fair share assumes that all flows are
  equal
• Different flows might have different QoS
  requirements
• e.g. voice and video flows require different
  bandwidth levels
• Customers of video service are willing to pay
  more to get the required bandwidth
• Associate a weight to each demand to take into
  account different requirements
• Normalize demands with corresponding weight
• Use weights to compute fair share

17
Max-Min Weighted Fair Share Example

• Demands d1 d2 10, d3 d4 25, d5 50
• Bandwidth C 100
• Max-Min satisfies d1, d2, d3, and d4 while d5
  gets 30
• If d5 has the same right as the others, use
  weighted Max-Min
• Weights w1 w2 w3 w4 1, w5 5
• Weighted demands d1 d2 10, d3 d4 25,
  d5 50 / 5 10
• FS C / (sum of weights) 100 / 9 11.11
• d1, d2 and d5 satisfied, d5 gets 50
• Sum of weights of unsatisfied demands M w3
  w4 2
• Bandwidth left R 100 (10 10 50) 30
• FS 30 / 2 15
• d3 and d4 not satisfied ? they get 15

18
Fair Queuing

• A router maintains multiple queues
• Separate queues for different traffic flows or
  classes
• Queues are serviced by a round-robin scheduler
• Empty queues are skipped over
• Bit-by-bit fairness among active flows
• Send one bit from every flow and repeat this
  process
• Packet-by-packet delivery in practice

1
2
...
K
19
Earliest Finish Time First

• For every packet i, calculate the finishing
  time(Fi) of packet transmission
• Fi Si Pi
• Si packet sending start time
• Pi number of clock ticks required to send
  packet i
• Si max(Fi-1, Ai), where Ai is the arrival time
  of packet i
• Clock increments when every active flow sends one
  bit
• Scheduler sends a packet with the lowest Fi
• Smaller packet that arrives late can be sent
  earlier than others
• Properties
• Work conserving
• No more than 1/n th share for any flow

20
Generalized Processor Sharing (GPS)

• Ideal fair queuing discipline based on fluid
  model
• Fluid model
• Traffic is infinitely divisible
• Can serve multiple flows at the same time
• Reflects weights
• Each nonempty queue is serviced for an
  infinitesimal amount of data in proportion to the
  associated weight
• Server capacity is split among nonempty queues
  according to weights
• Provides an exact Max-Min weighted fair share
  allocation
• WFQ is packet-by-packet GPS
• Lowest finish time first with weights

21
GPS Formal Definition

• GPS is a work conserving scheduler operating at
  fixed capacity C over K flows
• Each flow is characterized by a positive real
  weight
• is the amount of traffic from flow
  serviced in the interval
• GPS server is defined as a server for which
  for any flow that is backlogged (i.e. has
  nonempty queue) throughout

22
GPS Flow Rate

• Summing over all flows
• Each backlogged flow receives a service rate of
• If a flow source rate satisfies
  then such rate is always guaranteed independently
  of the other flows
• 0 lt N(t) ltK number of active flows at time t
• The service rate seen by flow when N(t) gt0
  iswith the sum taken over all active flows

23
Weighted Round Robin

• Each flow is characterized by its
• Weight
• Average packet size
• The server divides each flow weight by its mean
  packet size to obtain a normalized set of weights
• Normalized weights are used to determine the
  number of packets serviced from each connection
  in each round
• Simpler scheduler than WFQ
• Mean packet size may not be known a-priori

24
Deficit Round Robin

• DRR modifies WRR to handle variable packet sizes
  without knowing the mean packet size of each flow
  in advance
• DRR associates each flow with a deficit counter
• In each turn, the server attempts to serve one
  quantum" worth of bits from each visited flow
• A packet is serviced if the sum of the deficit
  counter and the quantum is larger than the packet
  size
• If the packet receives service, its deficit
  counter is reduced by the packet size
• If the packet does not receive service, a quantum
  is added to its deficit counter
• During a visit, if connection is idle, its
  deficit counter is reset to 0
• Prevents a connection from cumulating service
  while idle

25
Traffic Policing Shaping
26
Traffic Policing

• Source behavior must comply with traffic
  descriptor
• Traffic policing is performed at network edges to
  detect contract violations
• No queuing required
• Packets conforming to agreed bounds are forwarded
• Required resources are guaranteed
• Packets exceeding the agreed bounds can be
• Dropped at edge or
• Marked as non-conforming
• Dropped at any point in case of congestion
• No resource guarantee

27
Traffic Shaping

• In order to comply with descriptor, source
  traffic could be shaped to a predictable pattern
• Smoothing burstiness out
• Applied at source or network edges
• Exceeding packets are delayed
• Sent later when they eventually conform to
  descriptor
• Queuing required - overflow may cause
  loss/marking
• Latency introduced
• Traffic policing must still be enforced if
  shaping is left to the source

28
Traffic Policing vs. Traffic Shaping
Example based on peak rate
policing
Rate
Time
Peak rate
Rate
Rate
Time
shaping
Time
29
Traffic Shaping Leaky Bucket
data

• Purpose is to shape bursty traffic into a regular
  stream of packets
• Bucket is characterized by a size ß
• Flow is characterized by a drain rate, ?
• Impose a hard limit on transmission rate
• When bucket is full, packets can be
• Discarded
• Marked as non-conforming
• The effect of ß is to
• Limit the maximum bucket size
• Bound the amount of delay a packet can incur
• Given ß ? loss/marking rate vs. ? tradeoff
• ß 0 for peak rate policing

ß
?
30
Traffic Shaping Leaky Bucket

• Leaky bucket generates fixed-rate data flows
• QoS requirements easily guaranteed
• Suitable for smoothing small rate variations
• Depending on ß
• Highly variable rate sources must choose rate ?
  very close to their peak rate
• Wasteful solution
• Bursts are not permitted
• A shaper allowing limited rate variation at the
  output would be better

31
Traffic Shaping Token Bucket

• Bucket collects tokens
• Tokens are generated at rate ?
• Discarded when bucket is full
• Each packet requires a token to be sent
• A burst lesser than or equal to the number of
  tokens available can be transmitted (up to s)
• When bucket is empty, packets are buffered and
  sent at rate ?

?
tokens
s
data
ß
32
Traffic Shaping Token Bucket

• Number of packets sent in interval of length t
• N ? t s ? Linear Bounded Arrival Process
• ß 0 for LBAP policing
• Given ß, the minimal LBAP descriptor is not
  unique
• ? and s must be chosen
• Average rate A gt ? ? buffer grows without bound
  ? avoiding packet losses would require s to be
  infinite
• Peak rate P lt ? ? there are always tokens
  available ? s can be small at will
• As ? increases in the range A, P , the minimum
  s needed to meet the loss bounds decreases
• Any ? and its corresponding s is a minimal LBAP
  descriptor

33
Traffic Shaping Token Bucket
s
ß is fixed
s0
?0
?
A
P
34
Project Proposals
35
Sorted by Alphabetical Order

• Santiago Bock
• Jushua Geiger
• Sumedha Gupta
• Asim Jamshed
• Xuelian Long
• Victor Puchkarev
• Joshua Stachel
• Yinglin Sun
• Saman TaghaviZargar
• Pratik Vijay