DTIC ADA637177: A Feedback Implosion Suppression Algorithm for Satellite Reliable Multicast

7 Pages·0.1 MB·English

by  Defense Technical Information Center

T R R
ECHNICAL ESEARCH EPORT
A Feedback Implosion Suppression Algorithm for Satellite
Reliable Multicast
by Gun Akkor, John S. Baras, Michael Hadjitheodosiou
CSHCN TR 2003-3
(ISR TR 2003-8)
(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:3)(cid:6)(cid:7)(cid:3)(cid:8)(cid:4)(cid:9)(cid:10)(cid:8)(cid:4)(cid:11)(cid:12)(cid:7)(cid:3)(cid:13)(cid:13)(cid:14)(cid:7)(cid:3)(cid:4)(cid:12)(cid:6)(cid:15)(cid:4)(cid:16)(cid:17)(cid:18)(cid:8)(cid:14)(cid:15)(cid:4)(cid:5)(cid:10)(cid:19)(cid:19)(cid:20)(cid:6)(cid:14)(cid:21)(cid:12)(cid:7)(cid:14)(cid:10)(cid:6)(cid:4)(cid:22)(cid:3)(cid:7)(cid:23)(cid:10)(cid:8)(cid:24)(cid:25)(cid:4)(cid:14)(cid:25)(cid:4)(cid:12)(cid:4)(cid:22)(cid:26)(cid:11)(cid:26)(cid:27)(cid:25)(cid:28)(cid:10)(cid:6)(cid:25)(cid:10)(cid:8)(cid:3)(cid:15)(cid:4)(cid:5)(cid:10)(cid:19)(cid:19)(cid:3)(cid:8)(cid:21)(cid:14)(cid:12)(cid:13)(cid:4)(cid:11)(cid:28)(cid:12)(cid:21)(cid:3)
(cid:5)(cid:3)(cid:6)(cid:7)(cid:3)(cid:8)(cid:4)(cid:12)(cid:13)(cid:25)(cid:10)(cid:4)(cid:25)(cid:20)(cid:28)(cid:28)(cid:10)(cid:8)(cid:7)(cid:3)(cid:15)(cid:4)(cid:18)(cid:17)(cid:4)(cid:7)(cid:2)(cid:3)(cid:4)(cid:29)(cid:3)(cid:28)(cid:12)(cid:8)(cid:7)(cid:19)(cid:3)(cid:6)(cid:7)(cid:4)(cid:10)(cid:9)(cid:4)(cid:29)(cid:3)(cid:9)(cid:3)(cid:6)(cid:25)(cid:3)(cid:4)(cid:30)(cid:29)(cid:31)(cid:29) !(cid:4)(cid:14)(cid:6)(cid:15)(cid:20)(cid:25)(cid:7)(cid:8)(cid:17)!(cid:4)(cid:7)(cid:2)(cid:3)(cid:4)(cid:11)(cid:7)(cid:12)(cid:7)(cid:3)(cid:4)(cid:10)(cid:9)(cid:4)"(cid:12)(cid:8)(cid:17)(cid:13)(cid:12)(cid:6)(cid:15)!(cid:4)(cid:7)(cid:2)(cid:3)(cid:4)#(cid:6)(cid:14)$(cid:3)(cid:8)(cid:25)(cid:14)(cid:7)(cid:17)
(cid:10)(cid:9)(cid:4)"(cid:12)(cid:8)(cid:17)(cid:13)(cid:12)(cid:6)(cid:15)(cid:4)(cid:12)(cid:6)(cid:15)(cid:4)(cid:7)(cid:2)(cid:3)(cid:4)%(cid:6)(cid:25)(cid:7)(cid:14)(cid:7)(cid:20)(cid:7)(cid:3)(cid:4)(cid:9)(cid:10)(cid:8)(cid:4)(cid:11)(cid:17)(cid:25)(cid:7)(cid:3)(cid:19)(cid:25)(cid:4)&(cid:3)(cid:25)(cid:3)(cid:12)(cid:8)(cid:21)(cid:2)’(cid:4)(cid:1)(cid:2)(cid:14)(cid:25)(cid:4)(cid:15)(cid:10)(cid:21)(cid:20)(cid:19)(cid:3)(cid:6)(cid:7)(cid:4)(cid:14)(cid:25)(cid:4)(cid:12)(cid:4)(cid:7)(cid:3)(cid:21)(cid:2)(cid:6)(cid:14)(cid:21)(cid:12)(cid:13)(cid:4)(cid:8)(cid:3)(cid:28)(cid:10)(cid:8)(cid:7)(cid:4)(cid:14)(cid:6)(cid:4)(cid:7)(cid:2)(cid:3)(cid:4)(cid:5)(cid:11)(cid:16)(cid:5)(cid:22)
(cid:25)(cid:3)(cid:8)(cid:14)(cid:3)(cid:25)(cid:4)(cid:10)(cid:8)(cid:14)((cid:14)(cid:6)(cid:12)(cid:7)(cid:14)(cid:6)((cid:4)(cid:12)(cid:7)(cid:4)(cid:7)(cid:2)(cid:3)(cid:4)#(cid:6)(cid:14)$(cid:3)(cid:8)(cid:25)(cid:14)(cid:7)(cid:17)(cid:4)(cid:10)(cid:9)(cid:4)"(cid:12)(cid:8)(cid:17)(cid:13)(cid:12)(cid:6)(cid:15)’
(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:7)(cid:2)(cid:4)(cid:4)(cid:8)(cid:7)(cid:7)(cid:9)(cid:10)(cid:11)(cid:11)(cid:12)(cid:12)(cid:12)(cid:13)(cid:6)(cid:5)(cid:14)(cid:13)(cid:15)(cid:16)(cid:17)(cid:13)(cid:2)(cid:17)(cid:15)(cid:11)(cid:18)(cid:19)(cid:20)(cid:18)(cid:21)(cid:11)
Report Documentation Page Form Approved
OMB No. 0704-0188
Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it

does not display a currently valid OMB control number.
1. REPORT DATE 3. DATES COVERED
2003 2. REPORT TYPE 00-00-2003 to 00-00-2003
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
A Feedback Implosion Suppression Algorithm for Satellite Reliable
5b. GRANT NUMBER
Multicast
5c. PROGRAM ELEMENT NUMBER
6. AUTHOR(S) 5d. PROJECT NUMBER
5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
University of Maryland, College Park,Center for Satellite and Hybrid REPORT NUMBER
Communication Networks,Institute for Systems Research,College
Park,MD,20742
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)
11. SPONSOR/MONITOR’S REPORT
NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT
Approved for public release; distribution unlimited
13. SUPPLEMENTARY NOTES
14. ABSTRACT
In this paper, we propose a knapsack-based feedback suppression algorithm for reliable multicast
transport protocols operating over a satellite network. A reliable transport protocol needs to identify the
packets which failed to reach a given destination. This is achieved through feedback packets returned to
the source. For multicast services, receiver feedback has been shown to lead to the feedback implosion
problem. Feedback implosion is a well-studied problem and various solutions exist in the literature.
However, these solutions mainly focus on wireline terrestrial networks and do not take into account the
inherent characteristics of the satellite channel and the architecture of the deployed network. Therefore,
we need to revisit the problem and provide a new set of solutions for efficient integration to next generation
satellite systems. In this paper, we introduce a feedback implosion suppression algorithm, which effectively
suppresses the amount of feedback relayed through the satellite channel, while ensuring that the critical
information is conveyed in a timely fashion. The performance of the algorithm is evaluated through
simulations.
15. SUBJECT TERMS
16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF
ABSTRACT OF PAGES RESPONSIBLE PERSON
a. REPORT b. ABSTRACT c. THIS PAGE Same as 6
unclassified unclassified unclassified Report (SAR)
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std Z39-18
A Feedback Implosion Suppression Algorithm for
Satellite Reliable Multicast
Gun Akkor, John S. Baras, and Michael Hadjitheodosiou
Center for Satellite and Hybrid Communication Networks,
Institute for Systems Research, University of Maryland,
College Park, MD 20742, USA.
e-mail: {akkor, baras, michalis}@isr.umd.edu
Abstract—In this paper, we propose a knapsack-based feed- the flow of feedback packets from multicast receivers, which
back suppression algorithm for reliable multicast transport are locatedtypicallyat the leaves of the multicast distribution
protocolsoperatingover asatellitenetwork.Areliabletransport
tree, to the multicast sender, which is the root of the tree,
protocol needs to identify the packets which failed to reach a
causes network traffic concentration aroundthe links close to
given destination. This is achieved through feedback packets re-
turnedtothesource.Formulticastservices,receiverfeedbackhas the sender. Secondly, a sender is required to process a large
beenshowntoleadtothefeedback implosionproblem.Feedback number of feedback packets, which may be prohibitive in
implosionisawell-studiedproblemandvarioussolutionsexistin terms of memory, storage and computational load. However,
the literature. However, these solutions mainly focus on wireline
for satellite multicast services, there are other aspects to this
terrestrial networks and do not take into account the inherent
problem.Insatellite multicastnetworks,receiversarea single
characteristics of the satellite channel and the architecture of
the deployed network.Therefore, we need to revisit the problem hopawayfromthesatellite andthereis nophysicalhierarchy
and provide a new set of solutions for efficient integration to betweenthesatelliteandthereceivers.Therefore,themulticast
next generation satellite systems. In this paper, we introduce schemes that are based on physical or logical hierarchy for
a feedback implosion suppression algorithm, which effectively
aggregation of receiver feedback or local recovery cannot be
suppresses the amount of feedback relayed through the satellite
applied to this network topology [6]–[9]. As a result, the
channel,while ensuring that the critical information is conveyed
inatimelyfashion.Theperformanceofthealgorithmisevaluated feedback implosion problem becomes very challenging [10].
through simulations. Inasatellite topology,the problemis furtheraggravatedby
the fact that the return channel (uplink) is a shared medium
I. INTRODUCTION and the satellite spectrum resources are scarce and extremely
Broadband satellite systems are quickly becoming an inte- expensivecomparedto bandwidth on the ground.This means
gral componentof broadbandcommunicationnetworks. They that finding efficient solutions that keep feedback bandwidth
have several attractive characteristics, such as breadth of to a minimum value would preserve the available bandwidth
broadcast “reach”, ubiquitous access, low-cost global cover- forotheruseful traffic,andresult either in betterperformance
age,andastheymovetothenextgeneration,higherfrequency or in the ability to accommodatemoreusers, thus makingthe
band systems, they plan to offer large and flexible capacity, systemmorecompetitiveinpriceandperformance.Assigning
which would make them a natural technology option for a separate return channel to every multicast receiver would
carrying a variety of multicast services. However, despite the resultin the waste ofresourcesandcertainlywouldnotscale.
potential of satellite multicast, and the current market for Consideringthat (i) feedbackinformationmay containredun-
various satellite-based broadcast services, there exists little dant information (due to correlations among the loss pattern
support for reliable multicast via satellite. This is largely of receivers) and, (ii) most multicast algorithms only need
because most of the multicast protocols have been designed to track the behavior of a subset of receivers with the worst
primarilyforwirelineterrestrialnetworks,withouttakinginto case channel conditions, the challenge is to design efficient
accounttheinherentcharacteristicsofthesatellitechanneland algorithms to select and filter-out information from multiple
the deployed network topologies [1], [2]. receiverstoallowonlythemostrelevantfeedbackinformation
Many of the well-studied problems in wireline terrestrial to be conveyed to the source using as little bandwidth as
multicastwouldhaveeitherdifferentrootsordifferentsolution possible.
spacesinsatellitemulticast.Therefore,itisnecessarytorevisit In this paper, we address this issue by introducing a
someoftheseproblemsandprovideanewsetofsolutionsfor feedbackimplosion suppression algorithmthat allows the use
efficientintegrationofsatellitesystemswithmulticastservices of a fixed number ((cid:1) the number of receivers) of return
inthefuture.Forreliablemulticastprotocols,feedbackimplo- channelsbythemulticastgroupwhileensuringthatthecritical
sionisonesuchproblemthathasdrawnconsiderableattention information is conveyed to the satellite in a timely fashion.
fromtheresearchcommunity[2]–[5].Inthewirelineterrestrial The remainder of this paper is organized as follows. In
networks, feedback implosion occurs for two reasons. Firstly, Section II, we describe the generic reliable transport protocol
the receiver feedback on the EB, and proceeds with the
transmission of the next block.
• We assume that the network layers can detect corrupted
packetsanddiscardthem,i.ethereceiver’serrordetection
process is assumed to be perfect: an error-free packet is
neverrejected, and a corruptedpacket is never accepted.
Therefore,if the receivertransport layer receives at least
k·(1+(cid:1))uniqueencodingpacketsforaparticularEB,it
can decodethe block and pass the source blockto upper
layers. Otherwise, it buffers the uncorrupted packets of
the EB and creates a request for additional encoding
Fig.1. NetworkArchitecture packets. This request is the feedback transmitted to the
NOCanditidentifiestheEBandthenumberofadditional
encodingpacketsrequiredtodecodetheEB.Atanypoint
which operates between the source and the receivers. In intime,areceivermayhavebufferedseveralEB(s)which
Section III, we present our feedback suppression algorithm.
require additional encoding packets to complete. In this
SectionIVdescribesthesimulationenvironment,discussesthe
case,receiverssorttheirrequestsinapredeterminedorder
performancemetricsofinterest,andpresentsresults.SectionV
and transmit them one-by-one.
concludes the paper. • Anencodingpackettransmittedforaparticularencoding
blockbenefitsallreceiverswhichawaitadditionalpackets
II. ARELIABLEMULTICAST TRANSPORT PROTOCOL tocompletetheblock.Therefore,NOCwillbeinterested
in transmitting lmax = max{l1,...,lR} additional en-
Inordertoevaluatetheperformanceofafeedbacksuppres- coding packets for the EB, where lr is the number of
sionpolicy,weneedtodescribethebehavioroftheunderlying
additional encoding packets requested by receiver r for
reliable transport protocol which uses the receiver feedback
the EB, and R is the number of participating receivers
to detect packet losses and to initiate retransmissions. In this
(i.e. the number of receivers that have transmitted a
section,wedescribeagenericreliabletransportprotocolwhich
request for the particular block). NOC collects requests
operates between the source and the receivers. The protocol
from receivers and generates lmax additional encoding
descriptionis genericin thesense that it omits variousdetails
packets for the corresponding SB. These packets are
andpresentsonlythebehaviorrelevanttotheoperationofour
transmittedto the multicast groupeither usinga separate
policy.Severalvariationsofthisprotocolexistintheliterature
channel, or are piggy-backed with the next transmitted
and it has been shown to perform favorably in the context of
EB.
satellite multicast services [3], [11], [12].
In general, additional encoding packets may also get cor-
We consider a star network topology, where a geosyn-
ruptedduringthesubsequenttransmissionsandsomereceivers
chronous(GEO)satelliteprovidesmulticastservicestoalarge
may still need extra packets after the completion of the first
number of satellite users located inside its footprint. In this
request process. Therefore, receivers continue this repeat-
scenario, receiversare equippedwith two-way direct commu-
requestprocessuntiltheyaccumulateenoughencodingpackets
nication dishes, and access the terrestrial network through a
for decoding of the EB.
command center referred to as the network operations center
(NOC) (Figure 1). III. FEEDBACK SUPPRESSION POLICY
We assume the transport protocol performs as follows
When we consider the nature of the return information,
between the NOC (source) and satellite users (receivers):
we observe that, it is sufficient for the NOC to track only
• NOC organizes the source packets in source blocks the maximum number of encoding packets requested by any
(SB) of k packets and encodes them using a suitable receiver per encoding block. The volume of feedback would
expandable FEC code. An expandable FEC encoder be minimized if only the receiver with the maximum packet
takes k source packets as input and generates as many requirement responds to the NOC per encoding block. We
unique encoding packets as requested on demand. An can consider two extreme scenarios for this situation: (i) all
expandable FEC decoder has the property that any set receiverscommunicateamongeachotherthroughasecondary
of k · (1 + (cid:1)) unique encoding packets is sufficient to network(possiblya terrestrial connection)beforedecidingon
reconstruct the original k source packets, where (cid:1) is a whethertotransmitarequest,andsuppressthefeedbackofall
small decodingoverhead.Construction of an expandable receivers but the one with the maximum additional encoding
FEC code with small decoding overhead is described packet requirement; (ii) every receiver is assigned a separate
in [13]. The NOC initially generates n ≥ k unique return channel, which is used to communicate the request
encodingpacketscorrespondingto the packetsof the SB informationto the NOC, which then computesthe maximum.
and transmits the encoding block (EB) to the multicast The former scenario requires additional infrastructure and
group over the satellite. The NOC does not wait for collaboration between the receivers which are contradictory
to reasons for deployment of a satellite network. The latter not transmit a feedback packet when the condition changes
situationgivesrisetothefeedbackimplosionproblemaswell and it starts to build up a high average. Therefore, we must
as the waste of uplink resources. also allow receivers with previously low request averages to
Our solution strategy lies between these two boundaries transmit their requests in order to update their status at the
and attempts to minimize the number of transmitted request NOC.
messagesbythehelpofa knapsack-basedalgorithmthatruns Using these observations as the basis, the NOC solves
at the NOC without relying on any collaboration among re- the problem of deciding which receiver will be allowed to
ceivers.Weassumethatonlyafixednumberm(cid:1)Rofreturn transmitarequestinfutureuplinktransmissiontimes(because
channelsareallocatedtothemulticastsessionfortransmitting of the propagation delay, satellite can only calculate this
request packets. The problem is, then, to determine at every channel allocation layout for future times). We view each
uplink transmission time, which of the R multicast receivers of the available uplink channel as a knapsack with capacity
will be given a chance to transmit a request in one of the m ci = max{w1∗,...,wR∗} = c for i = 1,2,...,m and each
availablereturnchannels.Thisproblemfitsnicelytoamultiple receiver as an item with weight wr∗ and profit pr = wr∗ for
knapsackproblemformulation[14]whenoneconsidersevery r =1,...,R, where wr∗ is the last updatethe satellite has for
available return channel as a knapsack that needs to be filled theweightofreceiverr andRisthenumberofreceivers.We,
with items (receiver requests) of weight equal to the number then, solve the following multiple knapsack problem:
of additional packet requirement such that total weight is
maximized, i.e. receivers with maximum packet requirements (cid:1)m (cid:1)R
max p x
r ir
are assigned a channel for feedback transmission.
i=1r=1
We first make the following modifications to the receiver (cid:1)R
operation: subject to wr∗xir ≤ci =c for i=1...m (1)
r=1
• Instead of transmittinga requestimmediatelyin the next (cid:1)m (cid:1)R
x ≤1
uplink transmission time, which would be the case if ir
i=1r=1
every receiver is assigned a return channel, receivers
In the solution of this problem, xir = 1, if receiver r
buffer their requests in a queue and wait until they are
is assigned to return channel i. Once the channel allocation
assigned a feedback channel for transmission. In the
is determined, NOC broadcasts x and the value of c to all
meantime,ifadditionalencodingpacketsarereceived,the
receivers either using a separate control channel or by piggy-
pendingrequestsareupdatedtoreflect thenewnumbers.
backing it in the next EB. Upon receiving a new channel
A request for additional encoding packets can not be
allocation information, receivers apply it in the next uplink
for more than k (taking (cid:1) = 0 for simplicity) encoding
transmission time.
packets, at any time, since k unique encoding packets
It is important to note that the solution of this problem
suffice to complete the decoding of an EB.
may assign more than one receiver to a channel, while
• Every receiver r maintains a value wr, which is the some receivers are not assigned a channel in the next uplink
average number of additional encoding packets required transmission time. If a receiver has weight, w∗ that is closer
r
to complete all the pending encoding blocks. This in- to c, it is more likely to be assigned to a channelalone,since
formation is readily available at each receiver and can
itfillsthecapacityoftheknapsack(channel).Ahigherweight
be calculated simply by summing over the number of
indicatesthatthereceiverontheaverageneedsalargenumber
additional encoding packets required to complete all
of additional encoding packets to complete all the pending
pendingencodingblocksanddividingitbythenumberof
encoding blocks and hence its feedback will probably benefit
encodingblocks in the buffer. Clearly, 0≤wr ≤k. This the group the most. Therefore it is assigned to a channel
value is called the weight of that particular receiver and
alone. Receivers with smaller weights will be sharing the
issenttotheNOCalongwitharequestpacket,whenever
channel with other receivers. A smaller weight indicates that
the receiver is assigned a return channel.
thereceiverneeds,ontheaverage,onlyafewencodingpackets
• Receivers transmit the first request in their queue to the
to complete all the pending encoding blocks and hence its
NOC, whenever they are assigned a return channel.
feedback will only partially benefit the group. One may not
In general, receivers with higher packet request averages consider the receivers with small weights, but this is not a
should be given priority over other receivers since their re- good strategy since the algorithm uses the last updated value
quests would probably account for the requests of receiver of the receiver weights rather than the instantaneous value. A
with lower averages. This is due to the fact that an additional receiver with a small weight might build up a high average
encoding packet transmitted for an encoding block would sinceitslastupdate,andifwechoosetoassignonlythefirstm
benefit all receivers which have pending packets to complete receiverswiththehighestweightvaluestothereturnchannels,
that block. However, this would decrease the robustness of it may not get a chance to transmit a request and update its
the algorithm against instantaneous changes and variations in weight information.
the reception quality of receivers. Because, a receiver with a In order to avoid collisions, however, receivers do not
low packet request average is suppressed by its peers, it can transmit a repair request with probability equal to one in the
channel they are assigned to. They employ a probabilistic B. Simulation Environment
back-off algorithm and transmit a request with probability
In our simulations, we assume that the NOC has a fixed
proportionalto their current weights. We discuss the possible number B = 1000 of source blocks, each with k = 188
choices on the functional relation between the transmission packets.Eachsourcepacketis1Kbyte.Foreachsourceblock,
probability and receiver weights in Section IV-B. In the next the NOC constructs an encoding block of n = 204 encoding
section, we describe our simulation environment and present packets. There exists additional bandwidth of a=16 packets
simulation results on the performance of our algorithm.
for piggy-backing additional encoding packets in response to
receiver requests. The NOC and the receivers perform the
IV. SIMULATION RESULTS
previouslydescribedreliabletransportprotocoloverthesatel-
A. Channel Model lite channel described in Section IV-A. The choice of these
parameters affect the performance of the transport protocol
We modelthe satellite channelbya threshold-based2-state
but not the performance of our algorithm since we assume
MarkovChain.Inthismodelthechanneliseitherin“GOOD”
state, if the transmitted signal experiences less than Γ dB. thattheyareconstantwhethertheprotocolperformsfeedback
implosionsuppressionor not. Therefore,we assume that they
attenuation,or it is in “BAD” state, if the signal fade is more
than Γ dB., where Γ is the fade attenuation threshold [15], are fixed throughoutour simulations.
We need to define the relationship between the request
[16]. We assume that if the channel is in “GOOD” state,
channel coding is capable of correcting all bit errors and for transmitprobabilitypr(wr) of a receiverr andits weight wr.
simulationpurposestheprobabilityofbiterrorattheoutputof Intuitively, pr(wr) should satisfy pr(0) = a, pr(c) = 1 and
be a non-decreasing function of the receiver weight, where
the channel decoder is zero. In the “BAD” state, the channel
0≤a≤1. In this paper, we adopt:
behaves as a Binary SymmetricChannel (BSC) with bit error
probability equal to pb at the output of the channel decoder. exp(µ·w /c)
Thechannelstateischaracterizedbythetransitionprobability pr(wr)= exp(µ)r , (3)
matrix given by
(cid:2) r 1−r (cid:3) wherewr isthecurrentaverageofadditionalencodingpackets
M = , (2) receiverr wantstorequestandµisanaggressivenessparame-
1−s s
ter.Thisfunctionmaynotbethemostsuitablefunctionforthis
where r and 1−s are the probabilities that the channel will purpose and performance of the algorithm for different types
beinthe“GOOD”stateatthekth symbolduration,giventhat offunctionsiscurrentlyunderinvestigation.Inoursimulations
the channel was in “GOOD” and “BAD” states, respectively we set m
at the (k − 1)th symbol duration. Using the results of the µ=1− , (4)
R
ACTS PropagationExperiments,publishedin [17],for a fade
attenuationthresholdofΓ=10 dB., we haver =0.9999813, where m is the number of available return channels and
and 1−s=0.00172. R is the number of receivers. Therefore, receivers transmit
In orderto evaluatethe performanceof the transportproto- their requests more aggressively when more channels are
col,weneedtocalculatethebiterrorprobabilityattheoutput available, but back-off from transmitting as their average
of the channel decoder. In order to estimate the value of pb, weight decreases for a fixed number of return channels.
we use the link budget calculations of a commercial satellite
C. Performance Metrics
system proposed in [18]. The calculations confirm that for
signalattenuationoflessthan10dB.,linkbudgetmarginsand Thereare severalperformancemetricswhichmaybe ofin-
channelcodingarecapableofkeepingthebiterrorprobability terestinthisscenario.Inthispaper,weconsiderthefollowing
around 10−9 at the output of the channel decoder. For signal twometrics:(i)averagenumberofroundstocompleteablock,
attenuationof more than 10 dB., on the other hand, bit errors and(ii) bandwidthefficiency.Ifan encodingblockcan not be
occur with probability pb ≥ 0.1 at the decoder output of completed at the end of the first round, additional encoding
a typical concatenated channel encoder/decoder pair which packets need to be transmitted in the subsequent rounds. The
employs a RS(204,188) Reed-Solomon outer code and rate- firstmetricmeasurestheperformanceofthealgorithminterms
punctured inner convolutional codes (capable of supporting ofthe averagenumberofroundsit takes to completea block.
rates 1/2,2/3,3/4,5/6,7/8 to compensate for fading) [19], The second metric is the bandwidthefficiency in transmitting
[20]. Using this result, in our simulations, we evaluate the the original source packets and is given by the ratio of
time progress of the channel state for every bit transmission source block size to average number of packets transmitted
and assume that a bit is in error with probability pb = 0.1 tocompletetransmissionoftheblock.Otherpossiblemetrics,
whenthechannelisin“BAD”stateandnoerrorsoccurwhen which we would like to investigate in the future include
the channel is in “GOOD” state. Since these errors are at the encoding/decoding complexity and buffering requirements at
outputofthechanneldecoder,wefurtherassumethatapacket the NOC and the receivers. In the next section, we present
is corrupted if at least one bit is in error. We use this packet simulation results on the performance of our algorithm as a
loss pattern in our numerical results. function of the number of available return channels.
port protocols operating over a satellite network. We showed
that the suppression algorithm causes minimal degradationto
the performance of the underlying transport protocol while
effectively minimizing the volume of receiver feedback. We
alsoshowedthatitis possibletoimprovetheefficiencyofthe
transport protocol in terms of the overall bandwidth by more
conservative transmission of the additional encoding packets.
The majority of the multicast group may benefit from this
approach,whileonlyafewreceiverswithmaximumadditional
packet requirements are penalized. This paper presents a pre-
liminary set of results, and various other performance related
issues, such as the effect of variations in the loss pattern of
receivers,sizeofthereceiverset,choiceofparametersforthe
transport protocol,and the computationalload, are still under
Fig.2. Numberofpacketroundsversusnumberofavailablereturnchannels investigation.
forasetofR=100receivers.
REFERENCES
[1] F. Filali and W. Dabbous, “Issues on IP multicast service behaviour
over the next generation satellite-terrestrial hybrid networks,” in Proc.
ofComputerandCommunications, 2001,pp.417–424.
[2] M. W. Koyabe and G. Fairhurst, “Reliable multicast via satellite: A
comparison survey and taxonomy,” Int. J. Sat. Comm., vol. 19, no. 1,
pp.3–23,2001.
[3] J. Nonnenmacher, E. W. Biersack, and D. Towsley, “Parity-based loss
recovery for reliable multicast transmission,” IEEE/ACM Trans. Net-
working,vol.6,no.4,pp.349–361,August1998.
[4] C.Diot,W.Dabbous,andJ.Crowcroft,“Multipoint communication: A
surveyofprotocols,functions andmechanisms,” IEEEJ.Select. Areas
Commun.,vol.15,no.3,pp.277–290,1997.
[5] K.Obraczka, “Multicast transport protocols: Asurvey and taxonomy,”
IEEECommun.Mag.,January1998.
[6] J.P.Macker andW.K.Dang,“Themulticast dissemination protocol,”
NavalResearchLaboratory,Tech.Rep.,1996.
[7] S.Floyd,V.Jacobson,C.Liu,S.McCanne, andL.Zhang,“Areliable
multicast framework for light-weight sessions and application level
framing,”IEEEJ.Select.AreasCommun.,vol.15,pp.407–421,1997.
Fig. 3. Bandwidth efficiency as a function of number of available return [8] S.Paul,K.K.Sabnani,andD.M.Kristol,“Reliablemulticasttransport
channels forasetofR=100receivers. protocol (RMTP),”IEEEJ.Select. AreasCommun.,vol.15,no.3,pp.
407–421,April1997.
[9] T.Speakman,D.Farinacci,S.Lin,andA.Tweedly,“Pragmaticgeneral
multicast (PGM) transport protocol specification,” Internet Draft, June
D. Results
1999,workinprogress,draft-speakman-pgm-spec-03.txt.
In Figure 2, we plot the average number rounds it takes to [10] S.Cho and I. F.Akyildiz, “A poker-game-based feedback suppression
algorithmforsatellitereliablemulticast,”inProc.ofIEEEGLOBECOM
complete a block versus the number of available return chan-
2002,November2002.
nels for a receiver set of R = 100. We observe that average [11] S. Payne, “Reliable multicast via satellite,” Master’s thesis, University
number of rounds increases slightly as the number of return ofMaryland,CollegePark,MD,1999.
[12] K. Manousakis, “Reliable multicasting based on air caching for flat
channels decreases. However, the performance degradation is
heirarchy networks,” Master’s thesis, University of Maryland, College
minimal compared to the savings in terms of the number of Park,MD,2002.
returnchannels.InFigure3,weplotthe bandwidthefficiency [13] M.Luby,“Informationadditivecodegeneratoranddecoderforcommu-
nication systems,”U.S.PatnentNo.6,307,487,October2001.
versus the number of available return channels for the same
[14] D.Pisinger,“Anexactalgorithmforlargemultipleknapsackproblems,”
set. We observe that the bandwidth efficiency increases as EuropeanJournalofOperational Research,1999.
the number of return channels decreases. This is because the [15] E.O.Elliott,“Estimaesoferrorratesforcodesonburst-noisechannels,”
TheBellSystems,Tech.Rep.,1963.
feedbacksuppression algorithm does not always calculate the
[16] E. N. Gilbert, “Capacity of a busrt-noise channel,” The Bell Systems,
maximum,as it is the case when there is no suppression,and Tech.Rep.,1960.
on the average transmits at most as many packets as the case [17] J.SlackandM.Rice,“Finitestatemarkovmodelsforerrorburstsonthe
ACTS land mobile satellite channel,” NASA ACTS Experiment Paper
withoutthesuppression.This results in themoreconservative
ACTS-96-046,1996.
useofthebandwidth,butalsoincreasesthenumberofrounds [18] E. J. Fitzpatrick, “SPACEWAY system summary,” Space Communica-
it takes to complete a block as evident from Figure 2. tions,vol.13,pp.7–23,1995.
[19] “HUGHESDIRECWAYSystem,”www.direcway.com.
[20] J. Foerster and J. Liebetreu, “FEC performance of concatenated reed-
V. CONCLUSION solomon and convolutional coding with interleaving,” IEEE 802.16
Broadband WirelessAccessWorkingGroupReport,June2000.
In this paper, we introduced a knapsack-based feedback
implosion suppression algorithm for reliable multicast trans-