Tài liệu CDMA truy cập và chuyển mạch P5 doc

The multiple access scheme of the uplink control channel is based on aSpread Spectrum Random Access SSRA protocol, while the traffic channel access is an orthogonal CDMA scheme in both upl

In this chapter we provide channel assignment algorithms for optimum, optimum and random switch operation In each case, the system throughput hasbeen evaluated by simulation and the performance results are compared Performanceanalysis has been carried out for the case of optimum switch scheduling The analysis isbased on a discrete time Markovian model, and provides the call blockingprobabilitiesand data packet delays In Section 5.2 we describe the demand assignment system, andpresent the Traffic channel assignment control algorithms In Section 5.3 we providethe throughput analysis, and in Section 5.4 we present the performance results Thiswork originally appeared in reference [2].

The SS/CDMA demand assignment network is illustrated in Figure 5.1 The satellitehas an N × N Code Division Switch (CDS) and a Control Unit (CU) The interfacebetween the satellite and the SUs, called the Common Air Interface (CAI), consists

of control and traffic channels The control channels deliver signaling messages to andfrom the satellite while the traffic channels carry information data directly betweenthe end SUs The multiple access scheme of the uplink control channel is based on aSpread Spectrum Random Access (SSRA) protocol, while the traffic channel access is

an orthogonal CDMA scheme (in both uplink and downlink) called Spectrally EfficientCode Division Multiple Access (SE-CDMA) (see Chapters 3 and 6) Each SE-CDMA

Copyright © 2001 John Wiley & Sons Ltd ISBNs: 0-471-49184-5 (Hardback); 0-470-84169-9 (Electronic)

108 CDMA: ACCESS AND SWITCHING

ACRU

SBTU

CONTROL UNIT

NxN CODE DIVISION SWITCH

Traffic Channel

Signaling Channel

Figure 5.1 The SS/CDMA demand assignment network

frequency band (W ) can support up to L traffic channels The services assumed hereare both circuit switched and packet switched The circuit switched services are forvoice, video and data, while the packet switched service is only for data

The on-board switchingand control architecture is illustrated in Figure 4.2 ofChapter 4 Traffic channels are routed via the CDS module while signaling controlmessages are transmitted via the control channels, and are processed at the CU.The switchingsystem consists of Code Multiplexed Switch (CMS) modules EachCDS module routes calls between N uplink and N downlink beams within the samefrequency band, where each frequency band in each beam provides L traffic channels.The size of the CDS module then is N L× NL The CDS module system design hasbeen described in Chapter 4

The frequency band and the traffic channel allocations for circuit or packet switchedservices are made upon user request Message requests and assignments are sent via thecontrol channels, while the information data are transmitted via the traffic channels.This allows dynamic sharingof the available switchingresources between differenttypes of traffic While the intra-band switchingis performed by a single CDS mode, theintermodule or interband switchingof traffic will be handled by the demand assignmentmethod That is, given that the available spectrum consists of q pairs of uplink anddownlink bands, the module assignments are made upon call arrival This means thatwhile the control channel has a pre-assigned frequency band, the traffic channel band

is assigned to each SU upon call arrival, which tunes to it for transmitting and usethe correspondingmodule for routingits call

Circuit Calls have preemptive priority over data packetss

X : Circuit Calls

+

: Data Packets

Kc : Number of orthogonal Traffic Channels for circuit calls

Kp : Number of orthogonal Traffic Channels for data packets.

Wk : Orthogonal Traffic Channel codes k = 1, 2, …, K

be assigned momentarily for packets Traffic channels allocated for packet services arenot assigned for circuits Let kcbe the number of active circuit calls and kpthe number

of packets in transmission at a given time frame The traffic channel assignment ruleswill be based on the conditions:

if condition (b) does not hold true after the arrival of a new data packet, the packetwill be buffered

The assignments of circuit and packet switched services will be made out of a pool

of traffic channels In an uplink beam i, a traffic channel y(i)will be identified by the

110 CDMA: ACCESS AND SWITCHING

k

y

) i ( 1

y y(2i) y(3i) y(4i) ( i )

5

y y(6i)

) j ( 1

y y(2j) ( j )

3

y ( j ) 4

y y(5j) y(6j)

x Occupied Traffic Channels

)y,y(Y

)y,y(Y

) 4 ) 5 j , i 2

) 2 ) 3 ) , i 1

An end-to-end traffic channel (from uplink beam i to downlink beam j), Yk(i,j), isthen defined as an ordered pair of traffic channels Thus,

Yk(i,j)≡ (y(i)

k , yk(j)) for k = 1, , KThe timingdelay of the traffic channel assignment process is TA = w + 2tp, where2tp is the round trip propagation delay and w is the waitingtime for the assignment

to be made (All requests are kept on-board until a decision is made.) In the aboveequation, the assumption is that transmissions over the access channel (i.e the uplinkcontrol channel) is always successful Although the access channel transmissions have ahigh probability of success (0.9 or better), they are not always successful (The accesschannel will be designed to operate at a point of low-throughput and low-delay in

order to meet such a requirement.) Therefore, the actual delay is

TA= w + 2tp+ α(w + 2tp)where α is the average number of retransmissions required over the access channel.The system will also provide full duplex communication based on Frequency DivisionDuplexing(FDD) Each interbeam call requires the assignment of an uplink and adownlink band in each direction (a total of four bands) For intrabeam calls, the twouplink (or downlink) traffic channels may be separated either by frequency or by code.5.2.1 System Control Algorithm

The system model described here is based on a single CDS module (and thus onepair of frequency bands) of size N × N for switchingtraffic between N uplink and

N downlink beams Each beam has a capacity of L traffic channels in each frequencyband Lcand Lp traffic channels in each beam are used by circuit services and packetservices, respectively (Lc+ Lp= L) Any unused circuit channels can also be assignedmomentarily for packets However, traffic channels for data packets cannot be usedfor circuit calls

Duringeach frame, SUs send reservation requests for new circuit calls and datapackets to the CU via the uplink control channels The CU collects all requests inmatrix form, with rows and columns representingthe uplink and downlink beams,respectively Let Ta(k−1) and Da(k−1) be the circuit and data requests, respectively,

in frame (k− 1) The CU also maintains traffic matrix To(k− 1) of active (ongoing)call connections and matrix Db(k− 1) of unassigned (buffered) packet requests fromprevious frames, which are waitingin the SUs’ buffers to be assigned in subsequentframes In addition, at the end of the circuit call, the SU sends an indication to the

CU that its traffic channel becomes available, which is represented by Te(k− 1).Based on the g iven matrices in frame (k− 1), Ta(k− 1), Da(k− 1), Te(k− 1),

To(k−1) and Db(k−1), the CU applies an algorithm that makes assignment decisions

to determine matrices To(k), Tb(k), Do(k) and Db(k) for the next frame k To(k)represents the updated ongoing circuit calls, Tb(k) represents the blocked circuit calls,

Do(k) represents the assigned data packets, and Db(k) is the updated buffered datarequests The objective of the algorithm is to maximize the matrices To(k) and Do(k)for the given set of inputs The CU then passes the assignment decisions to the SUs viathe downlink control channels Each SU then transmits circuit calls and data packets

on the assigned traffic channels, while the CU provides the appropriate connections tothe CDS module Note that while the traffic channel is reserved for the entire duration

of a circuit call, each data packet transmission lasts only one frame

Figures 5.4-A and -B show the traffic flow for circuits and packets, respectively

In steady-state operation the flow equations for circuit calls and data packets can bewritten as:

T(k− 1) = To(k− 1) − Te(k− 1) + Ta(k− 1)T(k− 1) = To(k) + Tb(k)

D(k− 1) = Db(k− 1) + Da(k− 1)D(k− 1) = D (k) + D (k)

112 CDMA: ACCESS AND SWITCHING

kth Frame (k − 1)th Frame

Db(k

Da(k −1)

Db(k)

Da(k)

Da :New packet arrivals

Dr : Packets in the buffer to be scheduled in subsequent frames

D0 :Packets scheduled fo transmission

N

i=1

tij(To)≤ Lc for j = 1, , N (a− 1)N

j=1

tij(To)≤ Lc for i = 1, , N (a− 2)N

where the notation tij(X) stands for the (i, j) entry of matrix X Condition (a-1) saysthat the total number of traffic channels used for circuit calls destinated for downlinkbeam j cannot exceed Lc, and similarly, (a-2) restricts those originated from uplinkbeam i (b-1) and (b-2) restrict the total number of traffic channels used by bothcircuit calls and data packets to L.

Traffic Channel Assignment Algorithms

The Traffic Channel Assignment Algorithms (TCAAs) allow the CU to assign circuitcalls and data packets in order to achieve a high degree of channel utilization and meetcapacity constraints Three such algorithms are proposed here, TCAA-1 (Optimum),TCAA-2 (Fast/Sub-Optimum) and the Random Traffic Channel Assignment (RTCA)algorithm These algorithms are applied on matrices To(k−1), Te(k−1) and Ta(k−1)

of circuit calls and on matrices Db(k− 1) and Da(k− 1) of data packets of the currentframe, and provide the matrices To(k) and Tb(k) of circuit calls and the matrices

Do(k) and Db(k) of data packets for the next frame

In the description of the algorithms below, let

Traffic Channel Assignment Algorithm-1 (Optimum)

TCAA-1 utilizes a maximum flow algorithm to maximize the number of accepted calls

A bipartite graph is set up based on the number of traffic channels available and thenumber of new requests in each uplink and downlink beam The ‘maximum flow’ ofthat graph is computed which represents the requests that are accepted

Set up a network associated with A (see Remark 1)

Find the maximum flow in the network and the correspondingmatrix Ar (see Remark 2)

Step 1c:

Set To k) = Tr k−1)+Ar and Tb k) = Ta(k − 1) − Ar

114 CDMA: ACCESS AND SWITCHINGStep 2a:

Initialize matrix B = 0 Consider all (i, j) with

Set up a network associated with B Find the maximum

flow in the network and the corresponding matrix

Br.Step 2c:

3 TCAA-1 is optimal in the sense of maximizingthe number of accepted calls

or minimizingthe number of blocked calls This follows from the fact thatTCAA-1 is based on a maximal flow algorithm Matrix To(k) is maximized(i.e the sum of all the entries in the matrix is maximized) for given Ta(k−1)and Tr(k− 1) Similarly, Do(k) is maximized for given To(k) and Da(k− 1).Note, however, that the maximum flow, and hence To(k), provided in Step 1b

is not unique Therefore, further maximizingof Do(k) is possible if a differentmaximum flow or optimal matrix To(k) is used An example of TCAA-1 (forcircuit calls only) is given in Figure 5.5

Traffic Channel Assignment Algorithm-2 (Fast/Sub-Optimum)

TCAA-2 attempts to maximize the accepted calls in a forward blind manner byblockingnew calls that violate the schedulingconditions The matrix T(k− 1) =

Tr(k − 1) + Ta(k − 1) is reduced in each iteration of the algorithm until thecapacity constraints are satisfied After that, the iterations are repeated with matrixD(k− 1)

T T

| | | |

5 7 6 8

r c

i j

1 1 2

1

3 1 2 0

1 1 1 1

Find Matrix A=[a ij ] such that

Using the network associated with A, we find its maximum flow matrix A r

3 2 0 1

1 1 1 2

1

3 1 2 0

Step 2 of TCAA- operates on matrices T 0 and D in a similar

T

( c - [L )],

T

( r -

116 CDMA: ACCESS AND SWITCHINGStep 0:

Choose any row i with ri(Tm)> Lc; Goto Step 2a

If no such row exists, choose any column j with cj(Tm)>

If no such column exists, then choose any column j with

Step 5:

Choose any row i with ri(Dn+ To k)) > L; Goto Step 5a

If no such row exists, choose any column j with cj(Dn+

2 is a sub-optimum algorithm for providing traffic channel assignments

TCAA-2 has reduced computational complexity as compared to TCAA-1 (Its complexity inthe worst case is O(N2).) The algorithm attempts to maximize the circuit calls in Toalso considers maximizingthe traffic of data packets in Do (see Steps 2a and 2b).Random Traffic Channel Assignments (RTCA)

The Random Traffic Channel Assignment algorithm given below chooses the entries

in matrix Ta(k − 1) one by one randomly, and determines if that entry can beaccommodated The process is then repeated with Da(k− 1) This algorithm hasthe least possible computation complexity

118 CDMA: ACCESS AND SWITCHINGStep 4:

Time Versus Code Multiplexed Switching

Comparingthe time slot assignment problem in a Time Multipled Switch (TMS) withthe equivalent problem in a CDS, we can make the followingobservations In each timeslot of a TMS, in order to avoid conflict we must make no more than one assignment

to any input port i or output port j Now, given the assignments of ongoing callsover a frame, a new call request is said to be blocked if it results in conflict in everytime slot of the frame An optimum time slot assignment algorithm may then have torearrange the ongoing calls in order to accommodate the newly arrived calls One suchexample is shown in Figure 5.6 One of the newly arrived calls in the example wouldhave been blocked if rearrangement of the ongoing calls was not allowed It has beenshown [4] that for TMS, the throughput of an optimum assignment algorithm is 10%

to 15% higher than that of a random algorithm which does not allow rearrangement

of ongoing calls

In code switching, on the other hand, a new call request can be assigned to therequested input-output port (i.e uplink-downlink beam) as longas there are codes inthese beams available In other words, the requested route can be formed by pairingupany of the available codes of the input-output beams without rearranging the ongoingcalls Hence in code switching, optimization means collecting requests for a time period(a frame), and then assigning them so that the input-output flow is maximized, asdescribed by the above algorithms When this time frame becomes very small, theoptimum, sub-optimum and random assignments will converge If, for instance, onlyone new call request has arrived in the time frame, then there would be no differencebetween the three assignments These observations are verified by the results given inSection 5.4 under light traffic load

5.3 System Throughput Analysis

In this section we present the system performance analysis when the traffic channelassignment algorithm is optimum The analysis provides the blockingprobability forcircuit calls and the delay for data packets The performance of the optimum algorithmmay also be considered as the limitingcase of the sub-optimum or random algorithm,since the performance of the two systems will be the same when the number of newcall requests per frame is at most one This is often the case for circuit switched calls.When there is more than one request per frame, the performance analysis given here

is a tight bound of actual performance as obtained by simulation