allocation of layer bandwidths and fecs for video multicast over wired and wireless networks

Allocation of Layer Bandwidths and FECs forAllocation of Layer Bandwidths and FECs forVideo Multicast Over Wired and Wireless NetworksVideo Multicast Over Wired and Wireless Networks

T.-W. Angus Lee, S.-H. Gary Chan,

Qian Zhang, Wen-Wu Zhu, and

Ya-Qin Zhang

IEEE Trans. on CSVT. VOL. 12, NO. 12, DEC 2002

OutlineOutline

Introduction to “Layered Multicast”.Introduction to “Layered Multicast”.Architecture of the video systemArchitecture of the video system

Wired vs. wireless clients.Wired vs. wireless clients.Using a “transcoding” gateway or not.Using a “transcoding” gateway or not.

Recovery of packet-lossRecovery of packet-lossFeedback or forward-error correction code(FEFeedback or forward-error correction code(FE

C).C).

Architecture of the video Architecture of the video system.system.

Wired vs. wireless networksWired vs. wireless networks

In wired networksIn wired networksPackets are dropped mainly due to Packets are dropped mainly due to congestiocongestio

nn at the routers. at the routers.Packet-level FECPacket-level FEC in the form of in the form of parity parity packetspackets

are used to recover packet loss.are used to recover packet loss. In the wireless hopIn the wireless hop

Packets are often lost due to Packets are often lost due to random bit errorrandom bit errorss caused by fading of multipath effect. caused by fading of multipath effect.

Byte-level FECsByte-level FECs in the form of in the form of parity parity bytesbytes are are used to recover bit error.used to recover bit error.

Problems to be solvedProblems to be solved

Given the Given the heterogeneous errorheterogeneous error and and bandwidthbandwidth c characteristics of its end clients, how should the haracteristics of its end clients, how should the sserver allocateerver allocate the bandwidth and the correspondi the bandwidth and the corresponding ng packet-and-byte-level FECpacket-and-byte-level FEC of each video laye of each video layer in order to maximize the overall video quality?r in order to maximize the overall video quality?

Are there any differences in performance betweeAre there any differences in performance between a simple “n a simple “nontranscodingnontranscoding” gateway and the mo” gateway and the more complicated “re complicated “transcodingtranscoding” gateway?” gateway?

Measure of video qualityMeasure of video quality

PSNR (peak signal-to-noise ratio)PSNR (peak signal-to-noise ratio) Packet-loss rate after error correction has to be Packet-loss rate after error correction has to be

below a certain (low) value, e.g., below a certain (low) value, e.g., ≤≤1%1% for base la for base layer and ≤yer and ≤2%2% for enhancement layers. for enhancement layers.

The The PSNRPSNR is proportional to the video is proportional to the video goodputgoodput d defined as the efined as the useful data bits per seconduseful data bits per second (after e (after error correction) received by a client[15].rror correction) received by a client[15].

Maximizing the overall goodput.Maximizing the overall goodput.

Key issues in this paperKey issues in this paper HowHow FEC can be inroduced and applied in a video syste FEC can be inroduced and applied in a video syste

m vs. m vs. How muchHow much FECs is required, and other important is FECs is required, and other important issues shch as the optimal bandwidth of each layer and thsues shch as the optimal bandwidth of each layer and the value of a transcoding gateway.e value of a transcoding gateway.

Packet-levelPacket-level and and byte-levelbyte-level FECs should be combined fo FECs should be combined for optimal system operation.r optimal system operation.

A A receiver-driven multicast systemreceiver-driven multicast system with allocation at the with allocation at the sendersender..

Examined of layered multicast over Examined of layered multicast over mixed mediamixed media with with joijoint bandwidthnt bandwidth and and FEC allocationFEC allocation, and advantages of usin, and advantages of using a g a transcoding wireless gatewaytranscoding wireless gateway[32].[32].

Base-layer transmission andBase-layer transmission and its optimization its optimization

Allocated as the minimum end-to-end bit rAllocated as the minimum end-to-end bit rate.ate.

How much error control should be applied?How much error control should be applied?Quality is measured by the aggregate gooQuality is measured by the aggregate goo

dput or average goodput of the clients.dput or average goodput of the clients.

The FEC SchemesThe FEC Schemes

Nontranscoding gatewayNontranscoding gatewayPacket-level and byte-level FEC encoding are Packet-level and byte-level FEC encoding are

done at the video server, and error correction done at the video server, and error correction are only done at the end clients.are only done at the end clients.

The FEC generation schemeThe FEC generation scheme

Byte-Level:Byte-Level:Each symbol consists of m bits (m=8 in general)

Packet size: nb bytes

kb (≤1)bytes of source data packed with nb-kb parity bytes, where kb = nb, nb-2, ….

This is the so-called RS(nb, kb), which is able to correct up to tb symbol errors in

a packet, where tb = ceiling[ (nb – kb)/2 ].

The packet size nb is limited by 2m-1 symbols; therefore, for m=8, nb ≤255.

Packet-Level:Packet-Level:把 kp 個 byte-encoded video packets 中每個 packet 的每一個 byte 分別抽出來形成

nb 個大小為 np 的 packets. 比照 Byte-level 的做法，可以產生 np-kp 個 parity pac

kets.

因為 packet 是連續的，所以 tp=np-kp 個 packet losses 可以獲得修正。

The delay of the system will increase with np.

FormulationFormulation

To find out the optimal allocation between To find out the optimal allocation between the the video data ratevideo data rate, the , the packet-level FEC packet-level FEC raterate, and the , and the byte-level FEC ratebyte-level FEC rate..

Transcoding gatewayTranscoding gateway

Transcodes video packets from packet-levTranscodes video packets from packet-level FEC to byte-level FEC before forwardinel FEC to byte-level FEC before forwarding the packets to the wireless clients.g the packets to the wireless clients.

The FEC generation schemeThe FEC generation scheme

Quality OptimizationQuality Optimization

Optimization for Nontranscoding GOptimization for Nontranscoding Gatewayateway

Bit error rateBit error rateSymbol error rateSymbol error rateThe probability that a random packet The probability that a random packet

cannot be recovered by byte-level FEC is cannot be recovered by byte-level FEC is given bygiven by

For the wired clients, bit error rate is 0.For the wired clients, bit error rate is 0.The end-to-end packet drop rateThe end-to-end packet drop rate

Optimization for Nontranscoding GOptimization for Nontranscoding Gateway cont.ateway cont.

The probability that a random packet is perThe probability that a random packet is permanently “lost” is given by[6][7]manently “lost” is given by[6][7]

The goodput of the client g is hence given The goodput of the client g is hence given byby

Search on Search on kkpp and and kkbb

OO( Gn( Gnppnnb b ) => ) => OO( G(n( G(np p ++ nnbb)) ))For For kkpp, which the residual loss rate over th, which the residual loss rate over th

e wired netwoek is no more than e wired netwoek is no more than εεoo..

ααgg=0=0 for all clients. for all clients.

LetLetif ( ) { STOP and goto next step. }if ( ) { STOP and goto next step. }else { for all the clients with else { for all the clients with

search for the largest search for the largest }}

Search on Search on kkpp and and kkbb (cont.) (cont.)

For For kkbb , ,

Reintroduce Reintroduce ααgg for all wireless clients. Given for all wireless clients. Given kkpp**

such that such that εεgg for all the wireless clients are no mo for all the wireless clients are no mo

re than re than εεoo..

Optimization for Transcoding GateOptimization for Transcoding Gatewayway

The probability that a random packet is perThe probability that a random packet is permanently lost over the manently lost over the wired networkwired network is giv is given byen by

Optimization for Transcoding GateOptimization for Transcoding Gateway (cont.)way (cont.)

The end-to-end packet-loss rate after error The end-to-end packet-loss rate after error correction is given bycorrection is given by

and hence, the goodput of the clients isand hence, the goodput of the clients is

Compare the performance of transcoding anCompare the performance of transcoding and notranscoding gatewaysd notranscoding gateways

GG=10, half of them =10, half of them being wireless clients.being wireless clients.

Compare the performance of transcoding anCompare the performance of transcoding and notranscoding gateways (cont.)d notranscoding gateways (cont.)


es,g=0.18

Joint bandwidth and FEC optimization Joint bandwidth and FEC optimization for the enhancement layersfor the enhancement layers

The The base-layerbase-layer focuses mainly on focuses mainly on FEC FEC allocationallocation, the optimization of the , the optimization of the enhancement layers has two dimensions: enhancement layers has two dimensions: both both FEC and bandwidth allocationsFEC and bandwidth allocations..

The video is encoded into The video is encoded into LL enhancement enhancement layers. (layers. (LL+1+1 layers included the base layers included the base layer)layer)

A client cannot decode the layer A client cannot decode the layer ii without without receiving all of its preceding receiving all of its preceding ii-1-1 layers. layers.

Joint bandwidth and FEC optimization fJoint bandwidth and FEC optimization for the enhancement layersor the enhancement layers

Each of the layersEach of the layers is carried by a is carried by a multicast multicast groupgroup..

Assume that the video quality is enhanced Assume that the video quality is enhanced due to the enhancement layers is due to the enhancement layers is linearly linearly dependent ondependent on the the aggregate goodputaggregate goodput of th of the layers received.e layers received.

Optimization of the enhancement layersOptimization of the enhancement layers

What are the What are the bandwidthbandwidth and and FECFEC of of each each of the enhancement layersof the enhancement layers in order to maxi in order to maximize the mize the sum of video qualitysum of video quality enhanced in enhanced in terms of the terms of the goodput of each clientgoodput of each client..

Dynamic Program OptimizationDynamic Program Optimization

1. Ordering the end-to-end bandwidth in 1. Ordering the end-to-end bandwidth in increasing order.increasing order.


Cumulative transmission rate of the Cumulative transmission rate of the enhancement layers up to and including enhancement layers up to and including layer layer ll

We only need to considerWe only need to consider


LetLet SSll be the set of clients who join the be the set of clients who join the l-l-th enhanth enhan

cement layer, the sum of the good put for all the cement layer, the sum of the good put for all the clients joining enhancement layerclients joining enhancement layer l l is given by is given by


The total good put of the system due to the The total good put of the system due to the LL en enhancement layershancement layers

Denote the Denote the maximum goodputmaximum goodput ll enhancement layers enhancement layers

maximum end-to-end bandwidth is maximum end-to-end bandwidth is xx

Efficient Approximation on Layer Efficient Approximation on Layer BandwidthsBandwidths

In each of the recursive steps in the dynaIn each of the recursive steps in the dynamic program above, there are mic program above, there are O(O(GG)) possibi possibilities of lities of RR(l)(l). The search space of the above . The search space of the above bit rate and FEC allocation probem is bit rate and FEC allocation probem is O(O(GGLL((nnpp++nnbb))))..

Reduce the search space to Reduce the search space to O(O(LL((nnpp++nnbb)).)).

The ApproximationThe Approximation

Consider a large number of clients (i.e. Consider a large number of clients (i.e. GG→∞→∞ ) with their end-to-end bandwidth distri ) with their end-to-end bandwidth distributed according to some probability densitbuted according to some probability density function(pdf) y function(pdf) f(x)f(x) which ranges from which ranges from BBminmin t t

o o BBmaxmax..A total of clients are with A total of clients are with

enhancement bit rate of enhancement bit rate of

^ ^

The Approximation (cont.)The Approximation (cont.)

The aggregate goodput of all the clients foThe aggregate goodput of all the clients for the enhancement layersr the enhancement layers


Consider that the end-to-end bandwidth of Consider that the end-to-end bandwidth of the clients is uniformly distributed between the clients is uniformly distributed between BBminmin and and BBmax max ( with ( with mean(mean(BBmaxmax--BBminmin)/2)/2), i.e., ), i.e., ff

(x)(x) = 1/( = 1/(BBmaxmax--BBminmin)). Thus. Thus

^ ^ ^ ^

^ ^


The approximated layered bit rate is

End-to-end loss rate forEnd-to-end loss rate forEnhancement layers areEnhancement layers areBase-layer isBase-layer is

The client end-to-end bandwidths are The client end-to-end bandwidths are uniformly distributed betweenuniformly distributed between

the standard deviation isthe standard deviation is

Results of the joint bandwidth and Results of the joint bandwidth and FEC optimizationFEC optimization

Results of the joint bandwidth and Results of the joint bandwidth and FEC optimization (cont.)FEC optimization (cont.)

The and of each client are The and of each client are independent distributed with meanindependent distributed with mean

Transmission rate of enhancement layers Rl versus the standard deviation of the end-to-end bandwidth

of the clients

Average goodput versus the standard deviation of client bandwidth

Average goodput versus the standard deviatAverage goodput versus the standard deviation of client bandwidth for transcoding and nion of client bandwidth for transcoding and n

ontranscoding gatewaysontranscoding gateways

versus the standard deviation of client bandwidth

Average goodput versus the standard deviation of client bandwidth for different allocation strategies

Main contributionsMain contributions

A study of a video multicast system over wired A study of a video multicast system over wired and wireless networks with joint bandwidth and and wireless networks with joint bandwidth and FEC allocation for each layer in order to FEC allocation for each layer in order to maximize the over all video quality.maximize the over all video quality.

Present of an analytic model of the system, and Present of an analytic model of the system, and efficient algorithm on optimal FEC allocation for efficient algorithm on optimal FEC allocation for the base layer, and a dynamic program the base layer, and a dynamic program formulation with a fast and accurate formulation with a fast and accurate approximation on the optimal allocation of the approximation on the optimal allocation of the enhancement layersenhancement layers

Main contributionsMain contributions

Investigated of the advantages of using a Investigated of the advantages of using a gateway which transcodes from packet-levgateway which transcodes from packet-level FECs to byte-level FECs for the wirelesel FECs to byte-level FECs for the wireless link.s link.

allocation of layer bandwidths and fecs for video multicast over wired and wireless networks

Documents