Tài liệu Sổ tay của các mạng không dây và điện toán di động P8 ppt

In wireline networks, fluid fair queueing has long been a popular paradigm forachieving instantaneous fairness and bounded delays in channel access.. In flu- id fair queueing, during eac

CHAPTER 8

Fair Scheduling in Wireless

Packet Data Networks

THYAGARAJAN NANDAGOPAL and XIA GAO

Coordinated Science Laboratory, University of Illinois at Urbana-Champaign

8.1 INTRODUCTION

Recent years have witnessed a tremendous growth in the wireless networking industry.The growing use of wireless networks has brought the issue of providing fair wirelesschannel arbitration among contending flows to the forefront Fairness among users im-plies that the allocated channel bandwidth is in proportion to the “weights” of the users.The wireless channel is a critical and scarce resource that can fluctuate widely over a peri-

od time Hence, it is imperative to provide fair channel access among multiple contendinghosts In wireline networks, fluid fair queueing has long been a popular paradigm forachieving instantaneous fairness and bounded delays in channel access However, adapt-ing wireline fair queueing algorithms to the wireless domain is nontrivial because of theunique problems in wireless channels such as location-dependent and bursty errors, chan-nel contention, and joint scheduling for uplink and downlink in a wireless cell Conse-quently, the fair queueing algorithms proposed in literature for wireline networks do notapply directly to wireless networks

In the past few years, several wireless fair queueing algorithms have been developed[2, 3, 6, 7, 10, 11, 16, 19, 20, 22] for adapting fair queueing to the wireless domain In flu-

id fair queueing, during each infinitesimally small time window, the channel bandwidth isdistributed fairly among all the backlogged flows, where a flow is defined to be a logicalstream of packets between applications A flow is said to be backlogged if it has data totransmit at a given time instant In the wireless domain, a packet flow may experience lo-cation-dependent channel error and hence may not be able to transmit or receive data dur-ing a given time window The goal of wireless fair queueing algorithms is to make shortbursts of location-dependent channel error transparent to users by a dynamic reassignment

of channel allocation over small time scales Specifically, a backlogged flow f that ceives channel error during a time window [t1, t2] is compensated over a later time window

per-[t⬘1, t⬘2] when f perceives a clean channel Compensation for f involves granting additional channel access to f during [t⬘1, t⬘2] in order to make up for the lost channel access during

[t1, t2], and this additional channel access is granted to f at the expense of flows that were granted additional channel access during [t1, t2] while f was unable to transmit any data.

171

Handbook of Wireless Networks and Mobile Computing, Edited by Ivan Stojmenovic´

Copyright © 2002 John Wiley & Sons, Inc ISBNs: 0-471-41902-8 (Paper); 0-471-22456-1 (Electronic)

Essentially, the idea is to swap channel access between a backlogged flow that perceiveschannel error and backlogged flows that do not, with the intention of reclaiming the chan-nel access for the former when it perceives a clean channel The different proposals differ

in terms of how the swapping takes place, between which flows the swapping takes place,and how the compensation model is structured

Although fair queueing is certainly not the only paradigm for achieving fair and

bound-ed delay access in sharbound-ed channels, this chapter focuses exclusively on the models, cies, and algorithms for wireless fair queueing In particular, we explore the mechanisms

poli-of the various algorithms in detail using a wireless fair queueing architecture [15] In tion 8.2, we describe the network and wireless channel model, and give a brief introduc-tion to fluid fair queueing We also present a model for fairness in wireless data networks,and outline the major issues in channel-dependent fair scheduling In Section 8.3, we dis-cuss the wireless fair queueing architecture and describe the different policies and mecha-nisms for swapping, compensation, and achieving short-term and long-term fairness InSection 8.4, we provide an overview of several contemporary algorithms for wireless fairqueueing Section 8.5 concludes this chapter with a look at future directions

Sec-8.2 MODELS AND ISSUES

In this section, we first describe the network and channel model, and provide a briefoverview of wireline fluid fair queueing We then define a service model for wireless fairqueueing, and outline the key issues that need to be addressed in order to adapt fluid fairqueueing to the wireless domain

8.2.1 Network and Channel Model

The technical discussions presented in this chapter are specific to a packet cellular work consisting of a wired backbone and partially overlapping wireless cells Other wire-less topologies are briefly discussed in Section 8.5 Each cell is served by a base stationthat performs the scheduling of packet transmissions for the cell (see Figure 8.1) Neigh-boring cells are assumed to transmit on different logical channels All transmissions areeither uplink (from a mobile host to a base station) or downlink (from a base station to amobile host) Each cell has a single logical channel that is shared by all mobile hosts in thecell (This discussion also applies to multi-channel cellular networks, under certain re-strictions.) Every mobile host in a cell can communicate with the base station, though it isnot required for any two mobile hosts to be within range of each other Each flow of pack-ets is identified by a <host, uplink/downlink flag, flow id> triple, in addition to otherpacket identifiers

net-The distinguishing characteristics of the model under consideration are:

앫 Channel capacity is dynamically varying

앫 Channel errors are location-dependent and bursty in nature [5]

앫 There is contention for the channel among multiple mobile hosts

앫 Mobile hosts do not have global channel status (in terms of which other hosts arecontending for the same channel, etc.)

앫 The scheduling must take care of both uplink and downlink flows

앫 Mobile hosts are often constrained in terms of processing power and battery powerThus, any wireless scheduling and channel access algorithm must consider the constraintsimposed by this environment

In terms of the wireless channel model, we consider a single channel for both uplinkand downlink flows, and for both data and signaling Even though all the mobiles and thebase station share the same channel, stations may perceive different levels of channel errorpatterns due to location-dependent physical layer impairments (e.g., cochannel interfer-ence, hidden terminals, path loss, fast fading, and shadowing) User mobility also results

in different error characteristics for different users In addition, it has been shown in [5]

8.2 MODELS AND ISSUES 173

(Scheduler) Base Station

that errors in wireless channels occur in bursts of varying lengths Thus, channel errors arelocation-dependent and bursty This means that different flows perceive different channelcapacities Note that channel errors result in both data loss and reduce channel capacity.Although data loss can be addressed using a range of techniques, such as forward errorcorrection (FEC), the important issue is to address capacity loss, which is the focus of allwireless fair queueing algorithms

A flow is said to perceive a clean channel if both the communicating endpoints ceive clean channels and the handshake can take place A flow is said to perceive a dirtychannel if either endpoint perceives a channel error We assume a mechanism for the (pos-sibly imperfect) prediction of channel state This is reasonable, since typically channel er-rors, being bursty, are highly correlated between successive slots Hence, every host canlisten to the base station, and the base station participates in every data transmission bysending either data or an acknowledgement Thus, every host that perceives a clean chan-nel must be able to overhear some packet from the base station during each transmission

per-We assume that time is divided into slots, where a slot is the time for one completepacket transmission including control information For simplicity of discussion, we con-sider packets to be of fixed size However, all wireless fair queueing algorithms can han-dle variable size packets as well Following the popular CSMA/CA paradigm [9], we as-sume that each packet transmission involves a RTS-CTS handshake between the mobilehost and the base station that precedes the data transmission Successful receipt of a datapacket is followed by an acknowledgement At most one packet transmission can be inprogress at any time in a cell

Note that although we use the CSMA/CA paradigm as a specific instance of a wirelessmedium access protocol, this is not a requirement in terms of the applicability of the wire-less fair queueing algorithms described in this chapter The design of the medium accessprotocol is tied very closely to that of the scheduler; however, the issues that need to beaddressed in the medium access protocol do not limit the generality of the issues that need

to be addressed in wireless fair queueing [10, 11] The design of a medium access protocol

is a subject requiring detailed study and, in this chapter, we will merely restrict our tion to the impact a scheduling algorithm has on the medium access protocol

atten-8.2.2 Fluid Fair Queueing

We now provide a brief overview of fluid fair queueing in wireline networks Consider a

unidirectional link that is being shared by a set F of data flows Consider also that each flow f 僆F has a rate weight r f At each time instant t, the rate allocated to a backlogged flow f is r f C(t)/⌺i 僆B(t) r i , where B(t) is the set of nonempty queues and C(t) is the link ca- pacity at time t Therefore, fluid fair queueing serves backlogged flows in proportion to their rate weights Specifically, for any time interval [t1, t2] during which there is no

change in the set of backlogged flows B(t1, t2), the channel capacity granted to each flow i,

W i (t1, t2), satisfies the following property:

᭙i, j 僆 B(t1, t2), 冨 – ᎏW j (t1, t2)冨= 0 (8.1)

r

W i (t1, t2)ᎏ

r

The above definition of fair queueing is applicable to both channels with constant

capaci-ty and channels with time varying capacicapaci-ty

Since packet switched networks allocate channel access at the granularity of packetsrather than bits, packetized fair queueing algorithms must approximate the fluid model

The goal of a packetized fair queueing algorithm is to minimize |W i (t1, t2)/r i – W j (t1, t2)/r j|

for any two backlogged flows i and j over an arbitrary time window [t1, t2] For example,weighted fair queueing (WFQ) [4] and packet generalized processor sharing (PGPS) [18]are nonpreemptive packet fair queueing algorithms that simulate fluid fair queueing andtransmit the packet whose last bit would be transmitted earliest according to the fluid fairqueueing model

In WFQ, each packet is associated with a start tag and finish tag, which correspond spectively to the “virtual time” at which the first bit of the packet and the last bit of thepacket are served in fluid fair queueing The scheduler then serves the packet with the

re-minimum finish tag in the system The kth packet of flow i that arrives at time A( p i k) is

al-located a start tag, S( p i k ), and a finish tag, F( p i k), as follows:

S( p i k ) = max{V [A( p i k )], F( p i k–1)}

where V(t), the virtual time at time t, denotes the current round of service in the

corre-sponding fluid fair queueing service

F( p i k ) = S( p i k ) + L i k /r i where L i k is the length of the kth packet of flow i.

The progression of the virtual time V(t) is given by

=

where B(t) is the set of backlogged flows at time t As a result of simulating fluid fair

queueing, WFQ has the property that the worst-case packet delay of a flow compared tothe fluid service is upper bounded by one packet A number of optimizations to WFQ, in-cluding closer approximations to the fluid service and reductions in the computationalcomplexity, have been proposed in literature (see [22] for an excellent survey)

8.2.3 Service Model for Fairness in Wireless Networks

Wireless fair queueing seeks to provide the same service to flows in a wireless ment as traditional fair queueing does in wireline environments This implies providingbounded delay access to each flow and providing full separation between flows Specifi-cally, fluid fair queueing can provide both long-term fairness and instantaneous fairnessamong backlogged flows However, we show in Section 8.2.4 that in the presence of loca-tion-dependent channel error, the ability to provide both instantaneous and long-term fair-ness will be violated Channel utilization can be significantly improved by swappingchannel access between error-prone and error-free flows at any time, or by providing error

correction (FEC) in the packets This will provide long-term fairness but not neous fairness, even in the fluid model in wireless environments Since we need to com-promise on complete separation (the degree to which the service of one flow is unaffected

instanta-by the behavior and channel conditions of another flow} between flows in order to prove efficiency, wireless fair queueing necessarily provides a somewhat less stringentquality of service than wireline fair queueing

im-We now define the wireless fair service model that wireless fair queueing algorithmstypically seek to satisfy, and defer the discussion of the different aspects of the servicemodel to subsequent sections The wireless fair service model has the following proper-ties:

앫 Short-term fairness among flows that perceive a clean channel and long-term ness for flows with bounded channel error

fair-앫 Delay bounds for packets

앫 Short-term throughput bounds for flows with clean channels and long-term put bounds for all flows with bounded channel error

through-앫 Support for both delay-sensitive and error-sensitive data flows

We define the error-free service of a flow as the service that it would have received atthe same time instant if all channels had been error-free, under identical offered loads Aflow is said to be leading if it has received channel allocation in excess of its error-freeservice A flow is said to be lagging if it has received channel allocation less than its error-free service If a flow is neither leading nor lagging, it is said to be “in sync,” since itschannel allocation is exactly the same as its error-free service If the wireless schedulingalgorithm explicitly simulates the error-free service, then the lead and lag can be easilycomputed by computing the difference of the queue size of a flow in the error-free serviceand the actual queue size of the flow If the queue size of a flow in the error-free service islarger, then the flow is leading If the queue size of a flow in the error-free service issmaller, then the flow is lagging If the two queue sizes are the same, then the flow is insync

8.2.4 Issues in Wireless Fair Queueing

From the description of fair queueing in wireline networks in Section 8.2.2 and the scription of the channel characteristics in Section 8.2.3, it is clear that adapting wirelinefair queueing to the wireless domain is not a trivial exercise Specifically, wireless fairqueueing must deal with the following issues that are specific to the wireless environment

de-앫 The failure of traditional wireline fair queueing in the presence of dent channel error

location-depen-앫 The compensation model for flows that perceive channel error: how transparentshould wireless channel errors be to the user?

앫 The trade off between full separation and compensation, and its impact on fairness

of channel access

앫 The trade-off between centralized versus distributed scheduling and the impact onmedium access protocols in a wireless cell.

앫 Limited knowledge at the base stations about uplink flows: how does the base tion discover the backlogged state and arrival times of packets at the mobile host?

sta-앫 Inaccuracies in monitoring and predicting the channel state, and its impact on the fectiveness of the compensation model

ef-We now address all of the issues listed above, except the compensation model for flowsperceiving channel error, which we describe in the next section

Consider three backlogged flows during the time interval [0, 2] with r1= r2= r3 Flow 1and flow 2 have error-free channels, whereas flow 3 perceives a channel error during thetime interval [0, 1) By applying equation (1.1) over the time periods [0, 1) and [1, 2], wearrive at the following channel capacity allocation:

be inconsistent with fairness over a different time interval, though both time intervals havethe same backlogged set

In the fluid fair queueing model, when a flow has nothing to transmit during a time

window [t, t + ⌬], it is not allowed to reclaim the channel capacity that would have been allocated to it during [t, t + ⌬] if it were backlogged at t However, in a wireless channel, it

may happen that the flow is backlogged but unable to transmit due to channel error Insuch circumstances, should the flow be compensated at a later time? In other words,should channel error and empty queues be treated the same or differently? In particular,

consider the scenario when flows f1and f2are both backlogged, but f1perceives a channel

error and f2perceives a good channel In this case, f2will additionally receive the share of

the channel that would have been granted to f1 in the error-free case The question is

whether the fairness model should readjust the service granted to f1and f2in a future time

window in order to compensate f1 The traditional fluid fair queueing model does not need

to address this issue since in a wireline model, either all flows are permitted to transmit ornone of them is

8.2 MODELS AND ISSUES 177

In order to address this issue, wireless fair queueing algorithms differentiate between anonbacklogged flow and a backlogged flow that perceives channel error A flow that is notbacklogged does not get compensated for lost channel allocation However, a backlogged

flow f that perceives channel error is compensated in future when it perceives a clean

channel, and this compensation is provided at the expense of those flows that received

ad-ditional channel allocation when f was unable to transmit Of course, this compensation

model makes channel errors transparent to the user to some extent, but only at the expense

of separation of flows In order to achieve a trade-off between compensation and tion, we bound the amount of compensation that a flow can receive at any time Essential-

separa-ly, wireless fair queueing seeks to make short error bursts transparent to the user so thatlong-term throughput guarantees are ensured, but exposes prolonged error bursts to theuser

Exploring the trade-off between separation and compensation further, we illustrate a

typi-cal scenario and consider several possible compensation schemes Let flows f1, f2, and f3

be three flows with equal weights that share a wireless channel Let f1perceive a channel

error during a time window [0, 1), and during this time window, let f2receive all the

addi-tional channel allocation that was scheduled for f1(for example, because f2has packets to

send at all times, while f3has packets to send only at the exact time intervals determined

by its rate) Now, suppose that f1perceives a clean channel during [1, 2] What should thechannel allocation be?

During [0, 1], the channel allocation was as follows:

W1[0, 1) = 0, W2[0, 1) =2, W3[0, 1) = 1

Thus, f2 received one-third units of additional channel allocation at the expense of f1,

while f3received exactly its contracted allocation During [1, 2], what should the channelallocation be? In particular, there are two questions that need to be answered:

1 Is it acceptable for f3 to be impacted due to the fact that f1is being compensated

even though f3did not receive any additional bandwidth?

2 Over what time period should f1be compensated for its loss?

In order to provide separation for flows that receive exactly their contracted channel

allo-cation, flow f3should not be impacted at all by the compensation model In other words,the compensation should only be between flows that lag their error-free service and flowsthat lead that error-free service, where error-free service denotes the service that a flowwould have received if all the channels were error-free

The second question is how long it takes for a lagging flow to recover from its lag Of

course, a simple solution is to starve f2in [1, 2] and allow f1to catch up with the followingallocation:

W[1, 2] = 2, W [1, 2] = 0, W[1, 2) = 1

However, this may end up starving flows for long periods of time when a backlogged flowperceives channel error for a long time Of course, we can bound the amount of compen-sation that a flow can receive, but that still does not prevent pathological cases in which asingle backlogged flow among a large set of backlogged flows perceives a clean channelover a time window, and is then starved out for a long time until all the other lagging flowscatch up In particular, the compensation model must provide for a graceful degradation ofservice for leading flows while they give up their lead.

In a cell, hosts are only guaranteed to be within the range of the base station and not otherhosts, and all transmissions are either uplink or downlink Thus, the base station is theonly logical choice for the scheduling entity in a cell, making the scheduling centralized.However, although the base station has full knowledge of the current state of each down-link flow (i.e., whether it is backlogged, and the arrival times of the packets), it has limitedand imperfect knowledge of the current state of each uplink flow In a centralized ap-proach, the base station has to rely on the mobile hosts to convey uplink state informationfor scheduling purposes, which adds to control overhead for the underlying medium ac-cess protocol

In a distributed approach, every host with some backlogged flows (including the basestation) will have imperfect knowledge of other hosts’ flows Thus, the medium accessprotocol will also have to be decentralized, and the MAC must have a notion of priorityfor accessing the channel based on the eligibility of the packets in the flow queues atthat host (e.g., backoffs) Since the base station does not have exclusive control over thescheduling mechanism, imprecise information sharing among backlogged uplink anddownlink flows will result in poor fairness properties, both in the short term and in thelong term

In our network model, since the base station is involved in every flow, a centralizedscheduler gives better fairness guarantees than a distributed scheduler All wireless fairscheduling algorithms designed for cellular networks follow this model Distributedschedulers, however, are applicable in different network scenarios, as will be discussed inSection 8.5 The important principle here is that the design of the medium access control(MAC) protocol is closely tied to the type of scheduler chosen

When the base station is the choice for the centralized scheduler, it has to obtain the state

of all uplink flows to ensure fairness for such flows As discussed above, it is impossiblefor the centralized scheduler to have perfect knowledge of the current state for every up-link flow In particular, the base station may not know precisely when a previously non-backlogged flow becomes backlogged, and the precise arrival times of uplink packets inthis case The lack of such knowledge has an impact on the accuracy of scheduling and de-lay guarantees that can be provided in wireless fair queueing

This problem can be alleviated in part by piggybacking flow state on uplink sions, but newly backlogged flows may still not be able to convey their state to the basestation For a backlogged flow, the base station only needs to know if the flow will contin-

transmis-ue to remain backlogged even after it is allocated to a channel This information can be

8.2 MODELS AND ISSUES 179

easily obtained by the base station by adding a one bit field in the packet header For anonbacklogged flow, the base station needs to know precisely when the flow becomesbacklogged As far as we know, there exists no way to guarantee up-to-date flow state foruplink flows at the base station except for periodic polling, which may be wasteful interms of consuming excessive signaling bandwidth In related work [10, 11], two alterna-tive mechanisms are proposed for a base station to obtain this information, but thesemechanisms do not guarantee that the base station will indeed obtain the precise state ofuplink flows

Perfect channel-dependent scheduling is only possible if the scheduler has accurate mation about the channel state of each backlogged flow The location-dependent nature ofchannel error requires each backlogged flow to monitor its channel state continuously,based on which the flow may predict its future channel state and send this information tothe scheduler In CDMA cellular networks, a closed power-control loop provides the sig-nal gain for a host to the base station, accurate to a few milliseconds However, this maynot be sufficient for error bursts of a shorter duration In order to complement channelstate monitoring techniques, we need to predict the channel state based on previous knownstate, in a fairly accurate manner

infor-Errors in the wireless channel typically occur over bursts and are highly correlated insuccessive slots, but possibly uncorrelated over longer time windows [5] Thus, fairly ac-

curate channel prediction can be achieved using an n-state Markov model In fact, it has been noted that even using a simple one step prediction algorithm (predict slot i + 1 is good if slot i is observed to be good, and bad otherwise) results in an acceptable first cut

solution to this problem [11]

A wireless fair scheduler needs precise state information to provide tight fairness antees to flows If the scheduler has perfect state information, it can try to swap slots be-tween flows and avoid capacity loss However, if all flows perceive channel errors or thescheduler has imperfect channel state, then capacity loss is unavoidable In this sense,wireless fair queueing algorithms do not make any assumptions about the exact errormodel, though they assume an upper bound on the number of errors during any time win-

guar-dow of size T i , i.e., flow i will not perceive more than e ierrors in any time window of size

T i , where e i and T i are per-flow parameters for flow i The delay and throughput properties

that are derived for the wireless fair queueing algorithms are typically

“channel-condi-tioned,” i.e conditioned on the fact that flow i perceives no more than e ierrors in any time

window of size T i[10, 11]

8.3 WIRELESS FAIR QUEUEING ARCHITECTURE

In this section, we present a generic framework for wireless fair queueing, identify the keycomponents of the framework, and discuss the choices for the policies and mechanismsfor each of the components The next section provides instantiations of these choices withspecific wireless fair queueing algorithms from contemporary literature

Wireless fair queueing involves the following five components:

앫 Error-free service model: defines an ideal fair service model assuming no channelerrors This is used as a reference model for channel allocation

앫 Lead and lag model: determines which flows are leading or lagging their error-freeservice, and by how much

앫 Compensation model: compensates lagging flows that perceive an error-free nel at the expense of leading flows, and thus addresses the key issues of bursty andlocation-dependent channel error in wireless channel access

chan-앫 Slot queue and packet queue decoupling: allows for the support of both tive and error-sensitive flows in a single framework and also decouples connection-level packet management policies from link-level packet scheduling policies

delay-sensi-앫 Channel monitoring and prediction: provides a (possibly inaccurate) measurementand estimation of the channel state at any time instant for each backlogged flow

Figure 8.2 shows the generic framework for wireless fair queueing The different nents in the framework interact as follows The error-free service is used as the referencemodel for the service that each flow should receive Since a flow may perceive location-de-pendent channel error during any given time window, the lead and lag model specifies howmuch additional service the flow is eligible to receive in the future (or how much service theflow must relinquish in the future) The goal of wireless fair queueing is to use the compen-sation model in order to make short location-dependent error bursts transparent to the lag-ging flows while providing graceful service degradation for leading flows In order to sup-port both delay-sensitive and error-sensitive flows, the scheduler only allocates slots toflows and does not determine which packet will be transmitted when a flow is allocated aslot Finally, the channel prediction model is used to determine whether a flow perceives aclean or dirty channel during each slot (If the channel is dirty, we assume that the channelprediction model can also predict the amount of FEC required, if error correction is used.)Once a flow is allocated a slot, it still needs to perform the wireless medium access al-gorithm in order to gain access to the channel and transmit a packet We do not explore theinteractions between the scheduling algorithm and the medium access algorithm in thischapter We now proceed to describe the components of the architecture, except channelmonitoring and prediction, which have been described earlier

compo-8.3.1 Error-Free Service Model

The error-free service model provides a reference for how much service a flow should ceive in an ideal error-free channel environment As mentioned above, the goal of wirelessfair queueing is to approximate the error-free service model by making short error burststransparent to a flow, and only expose prolonged channel error to the flow

re-Most contemporary wireless fair queueing algorithms use well-known wireline fairqueueing algorithms for their error-free service model Three choices have typically beenused:

8.3 WIRELESS FAIR QUEUEING ARCHITECTURE 181

Figure 8.2 Generic framework for wireless fair queueing.

Is f1 in error ?

Is f1 leading ?

flow f2 to transmit algorithm and select

NOYES

YES

NO

YESYES

Transmit packet

Slot and Packet Queues Lead and Lag model