An Overview of Scheduling Algorithms in Wireless Multimedia Networks

1
An Overview of Scheduling Algorithms in
Wireless Multimedia Networks

• ???
• Hsiao-Chih George Lee
• http//www.en.oit.edu.tw/lee
• mailto//lee_at_en.oit.edu.tw

2
Abstract

• Scheduling algorithms are important components in
  the provision of guaranteed quality of service
  parameters such as delay, delay jitter, packet
  loss rate, or throughput.
• The design of scheduling algorithms for mobile
  communication networks is especially challenging
  given highly variable link error rates and
  capacities, and the changing mobile station
  connectivity.
• This article provides
• a survey of scheduling techniques for several
  types of wireless networks.
• Scheduling algorithms for TDMA, CDMA, and
  Multihop networks.

3
Introduction

• QoS requirements for different service classes

4

• A typical wireless scheduler

5
Wireless Network Scheduler Challenges

• Characteristics of wireless links
• Subject to time- and location-dependent signal
  attenuation, fading, interference, and noise that
  result in bursy errors and time-varying channel
  capacities.
• Wireless channel model
• Discrete-time Markov chain with two states
  error-free (good) or error-prone (bad)
• A packet is successfully received if and only if
  the link stays in the good state throughout the
  packet transmission time.

6

• Information needed to make scheduling decisions
• Number of sessions
• Session reserved rates
• Link states
• Statuses of session queues
• Information availability
• For the down-link
• The scheduler is located at the base station (BS)
• This information is easily obtianed
• For the up-link
• Some means must be provided to collect queue
  status information and to inform mobile stations
  (MSs) of their transmission times.

7

• To maximize MS battery life
• To transmit/receive in contiguous time slots and
  then go into a sleep mode rather than to rapidly
  switch among transmit, receive and sleep modes.
• Handoffs
• Following a handoff, any packets for S that are
  queued at previous cell C1s BS will be forward
  to current cell C2s BS
• For timestamp-based scheduling
• Timestamp update
• Fairness gap low timestamp ? extra service

8

• In CDMA network,
• the total interference at an MS must be small
  enough to ensure an adequate signal-to-interferenc
  e ration (SIR) for each session, thereby enabling
  its target bit error rate (BER) to be met.
• The scheduler must ensure that the number of
  simultaneous transmissions in the network is not
  so high as to result in excessive interference.
• In multihop networks
• No BSs
• Rapidly changing topology
• Routing

9
Scheduler Components and Properties

• Components
• (1) An error-free service model
• (2) A lead/lag counter
• Whether the session is leading, in sync with, or
  lagging its error-free model and by how much
• (3) A compensation model for each session
• A lagging session is compensated at the expense
  of leading sessions

10

• (4) Separate slot queues and session queues for
  each session
• When a packet arrives, it is timestamped and
  placed in the packet queue
• A slot with the same timestamp value is added to
  the slot queue.
• If the HOL (Head of line) packet for a session is
  dropped due either to excessive delay
  (delay-sensitive) or an excessive number of
  retransmissions (error-sensitive), the precedence
  of the session for accessing the channel is
  maintained by the slot queue.
• (5) A means for monitoring and predicting the
  channel state for every backlogged session.

11

• Features
• Efficient link utilization
• Delay bound
• Fairness
• Throughput
• Implementation complexity
• Graceful service degradation
• Isolation
• Energy consumption
• Delay/bandwidth decoupling
• Scalability

12
Scheduler Classification

• Work-conserving
• The scheduler is never idle if there is a packet
  awaiting transmission.
• Generalized Processor Sharing (GPS)Weighed Fair
  Queueing (WFQ)Virtual Clock (VC)Weighted Round
  Robin (WRR)Self-Clocked Fair Queueing (SCFQ)
  Deficit Round Robin (DRR)

13

• Non-work-conserving
• The scheduler may be idle even if there is a
  backlogged packet in the system because it may be
  expecting another higher-priority packet to
  arrive.
• Hierarchical Round-Robin (HRR) Stop-and-Go
  Queuing (SGQ)Jitter-Earliest-Due-Date
  (Jitter-EDD)
• ? Higher average packet delays than
  work-conserving
• ? May be used in application where jitter is more
  important than delay

14

• Timestamped
• Incoming packets are timestamped before being
  placed in their respective session queues.
• The HOL packets are then sorted in increasing
  order of their timestamps, and the packet with
  the lowest timestamp value is selected for
  transmission.
• Better QoS guarantees
• Round-robin
• No timestamps
• Easily implemented

15

• Sorted-priority
• Each session has a different priority level
• Packets are chosen for transmission according to
  their session priority
• VC, WFQ, Jitter-EDD
• Frame-based
• Time is divided into frames of fixed or variable
  size
• Each session reserves a portion of the frame for
  transmission
• Non-work-conserving
• HRR, SGQ fixed-size frames
• WRR, DRR, FFQ (Frame-based Fair Queueing)
  variable-size frames

16
Generalized Processor Sharing-Based Scheduling

• GPS
• An ideal fluid flow model that services all
  sessions simultaneously
• Work-conserving
• PGPS (Packet GPS)
• To emulate GPS by serving packets in increasing
  order of their timestamps

17
(No Transcript)
18
Scheduling Algorithms for Wireless TDMA Networks

• Network Model
• One BS and a number of MSs.
• Scheduling is implemented at the BS, which can
  communicate with all MSs
• Direct MS-to-MS communication is not possible.
• Channel errors.
• Knowledge of the channel state and packet queue
  status for all sessions is available at the BS.

19

• Channel State Dependent Packet Scheduling (CSDPS)
• To avoid bursty errors at the link layer
• Allow the use of different service disciplines
• RR
• Longest Queue First (LQF)
• Earliest Timestamp First (ETF)
• Channel state for each session is monitored.
• If the channel for a session is in a bad state,
  the transmission of this packet is deferred.
• No lead and lag
• No compensation

20

• Idealized Wireless Fair Queueing (IWFQ)
• Realization of PGPS with a compensation
• Error-free network model WF2Q
• Each session has a service tag that is equal to
  the virtual finish time of its HOL packet.
• Lagging sessions will have the lowest service tag
  values
• Bound on the lead and lag session
• Not graceful lagging session capture the channel
  until it has been compensated for all its lag

21

• Channel-Condition-Independent Fair
  Queueing(CIFQ)
• Use of Start-time Fair Queueing (SFQ)
• Scheduling is based on the start time
• Each session has a lead and lag counter
• Graceful service
• A leading session retains a fraction ? ? 0, 1
  of its service
• No lead/lag counter bound is needed

22

• Server-Based Fairness Approach (SBFA)
• A portion of the outgoing bandwidth is reserved
  for a hypothetical session called a Long-Term
  Fairness Server (LTFS).
• The LTFS is used to compensate lagging sessions.
• When a session is selected for transmission, it
  is not allowed to transmit if it has a bad
  channel, a slot for this session is created and
  queued into the LTFS session.
• No lead/lag counter
• The compensation rate depends on the reserved
  bandwidth for LTFS.

23

• Wireless Fair Service (WFS)
• Timestamp
• Sessions are selected for transmission in
  increasing order of their service tag values
  provided that the virtual start time of the HOL
  packet is less than v(t) ? where ? is a
  look-ahead parameter that determines the
  schedulable interval.
• If ? ? and Ri ?i, the error-free service is
  WFQ.
• If ? 0 and Ri ?i, the error-free service is
  WF2Q.
• Lead/lag counter

24

• Summary

25
(No Transcript)
26
Scheduling in CDMA Networks

• Advantages of CDMA over TDMA and FDMA
• higher (soft) system capacity
• soft handoff
• simple frequency planning
• inherent frequency diversity against multipath
  fading
• Voice activity factor and antenna sectorization
  are readily exploited using CDMA.
• Drawback
• an accurate power control mechanism is required.

27

• Properties of CDMA
• The soft capacity feature of CDMA allows a new
  session to be established provided that the for
  all transmitting sessions can be maintained above
  their target levels a certain percentage of the
  time 11.
• The packets sent from a number of MSs can be
  successfully received simultaneously at the BS,
  provided an adequate power control scheme is used.

28

• Packet-By-Packet GPS (PGPS)
• A residual power index that is not utilized.
• To find one or more additional (non-HOL packets)
  packets which can be transmitted.

29

• Scheduled CDMA (SCDMA)
• A hybrid CDMA/TDMA scheduler in which the BS
  schedules transmissions of the MSs as shown in
  Fig. 2.
• A fixed-size unit called capsule which can
  accommodate one or more packets.
• All MSs in a cell have the same capsule size.
• Assume a time-slotted operation
• For uplink scheduling
• (1)A capsule transmission request (CTR) is sent
  to BS by an MS
• (2) The BS sorts the CTRs according to their
  priority level or delay tolerance and places them
  in a global queue.
• (3) The BS sends transmission permission to the
  selected MS to inform of their capsule
  transmission times and power levels

30
?
?
?
31

• Dynamic Resource Scheduling (DRS)
• Centralized and adaptive
• MSs send their requests to the BS
• The BS classifies them according the traffic
  characteristics of the requested service and
  places them in two separate queues guaranteed
  queue (CBR, VBR, minimum-rate ABR) and
  best-effort queue (UBR and excess ABR)
• Neither DRS and SCDMA provide guaranteed delay
  bound

32
(No Transcript)
33

• Wireless Multimedia Access Control Protocol with
  BER Scheduling (WISPER)
• Packets are transmitted within frames of length T
  in both uplink and downlink,
• The uplink request slot is used by an MS to place
  transmission requests.
• The downlink control slot is used by the BS to
  grant transmission permissions.
• Traffic classes with a different BER requirement
• The selected packets are grouped into the N slots
  according to the following order
• Same BER?stricker BER?looser BER

34
(No Transcript)
35
Scheduling in Multihop Networks

• Network Model
• Base stations are not available
• Single-hop networks
• Each MS can communicate directly with all other
  MSs
• multi-hop networkMobile Ad-hoc Networks (MANETs)
• Not all MSs can communicate directly with each
  other.
• MSs, acting as relays, are then needed to forward
  packets to their final destinations.

36
Scheduling in Multihop Networks

• Base stations are not available
• Single-hop networks
• Each MS can communicate directly with all other
  MSs
• multi-hop networkMobile Ad-hoc Networks (MANETs)
• Not all MSs can communicate directly with each
  other.
• MSs, acting as relays, are then needed to
  forward packets to their final destinations.

37

• Network Model
• Time is assumed to be divided into slots, each of
  duration equal to one maximum-length packet
  transmission time plus the maximum propagation
  time between any two MSs.
• Omni-directional antenna
• Noise-free ? an unsuccessful reception is due
  only to collisions.
• Half-duplex?an MS can transmit or receive, but
  cannot do both at the same time.

38

• Conflict Types in Multihop Scheduling
• Two types of conflicts
• A primary conflict occurs when two or more MSs
  simultaneously transmit to a common destination
  node (e.g., nodes C ? D F ? D).
• A secondary conflict occurs when an MS receiving
  a transmission intended for it is interfered with
  by another transmission not intended for it
  (e.g., C? D F ? G)

Different spreading code ? a secondary conflict
can be tolerated
39

• Scheduler Properties
• Topology transparency
• The scheduler should work efficiently regardless
  of how frequently and unpredictably the topology
  changes.
• A topology independent algorithm reduces the
  burden of having to recompute and reassign time
  slots.
• Low connectivity information requirement
• Some algorithms need global network connectivity
  information while others require only local
  (e.g., one- or two-hop) connectivity information.
• Scheduler Examples
• Node activation scheduler
• Link activation scheduler

40

• (1) Node activation scheduler
• selects nodes for transmission to ensure that a
  packet transmission from any node will be
  received correctly by all its neighbors.
• Each MS has network connectivity information
  within a two-hop radius.
• Each MS i is allocated the ith time slot in a
  frame
• MSs more than two hops away from the MS i are
  also eligible to transmit during slot i.
• Some pre-established rule is used to select an
  eligible MS to transmit in slot i.
• The selected MS sends a broadcast message to
  inform other MSs that it is using slot i.
• It does not ensure fair slot allocations among
  all MSs and is not topology-transparent.

41

• Scheduler Examples
• (1) Node activation scheduler
• selects nodes for transmission to ensure that a
  packet transmission from any node will be
  received correctly by all its neighbors.
• Each MS has network connectivity information
  within a two-hop radius.
• Each MS i is allocated the ith time slot in a
  frame
• MSs more than two hops away from the MS i are
  also eligible to transmit during slot i.
• Some pre-established rule is used to select an
  eligible MS to transmit in slot i.
• The selected MS sends a broadcast message to
  inform other MSs that it is using slot i.
• It does not ensure fair slot allocations among
  all MSs and is not topology-transparent.

42

• (2) Link activation scheduler
• selects nodes for transmission to only guarantee
  that the destination node receives the packet
  successfully.
• MS i uses a time slot assignment function (TSAF)
  to locate its transmission slots in a frame.
• The TSAF, fi(x), is a polynomial of degree k with
  coefficients in 0, 1, , p 1 where p is a
  prime number.
• Polynomial evaluation is done using mod p
  arithmetic.
• Each frame is divided into p subframes, each
  consisting of p slots.
• MS i is assigned one slot in each subframe as
  follows in subframe j, j 0, 1, , p 1, the
  assigned slot number is given by fi(j).
• A collision occurs between MS i and MS l in
  subframe j if fi(j) fl(j).

43
Conclusion

• Much work remains to be done in the design and
  performance evaluation of wireless schedulers,
  especially for CDMA and multihop networks.