Wavelength/Waveband-Routed Networks in Software Draw QR Code JIS X 0510 in Software Wavelength/Waveband-Routed Networks

Wavelength/Waveband-Routed Networks use software qr code generation toconnect qrcode in software International Standard Serial Numbers how the random nature of dema Software QR Code ISO/IEC18004 nd-assigned traf c, together with its granularity, affects system performance. To model random activity in the network, we assume that many elementary stations, each with a single tunable transceiver, access each node. (The stations might access the nodes directly or through LASs.

) The station population is assumed to be suf ciently large so that the connection request arrivals can be approximated as a Poisson process. Assuming an LCC model, the blocking probabilities can then be deduced using the Erlang-B formula of Equation (5.91).

As we shall see, the blocking performance depends on the way channels are shared in the LLN. Let us assume that the collection of stations accessing each node generates an offered traf c G (Erlangs per node), where the destinations of the connection requests are distributed uniformly to N other nodes. (Any blocking because of busy destination stations is neglected.

) Suppose that one point-to-point optical connection is in place for each of the N outbound paths to the other nodes; that is, one -channel is available for each destination. In this case, each path is offered G/N Erlangs, so that the Erlang formula reduces to PB = G/(N + G), which means that the offered traf c must be kept extremely low for acceptable blocking probability. As an example, consider the Petersen network with W = 9 and C = 1, in which many elementary access stations are attached to each node possibly through LASs, as shown in Figure 6.

45. Each node generates G Erlangs, distributed uniformly to nine other nodes, with one -channel available to each destination. Then we nd that G 0.

38 for PB = 0.04. This corresponds to a total normalized carried load (throughput) for the network of S = 3.

6, compared with a maximum throughput of S = 90 in the deterministic case. The poor throughput for demand-assigned traf c is due to the fact that the number of channels is too small (the granularity is too large). A brute force way of improving PB is to assign many -channels (i.

e., many parallel optical connections) to each path. For example, taking C = 30 in the previous example, we nd that G 216 for PB = 0.

04. This gives a carried load of S = 2074, compared with S = 2700 in the deterministic case. Although this is much better, the random nature of the connection activity still degrades throughput by more than 20% with a nonnegligible blocking probability.

Unfortunately, in multiwavelength networks, large numbers of -channels may not be available. However, blocking probability can still be reduced using a limited number of -channels by re ning the granularity of the LCs. Suppose that instead of occupying a full -channel, each LC requires a lower effective bit rate: Rt /K , where K 1.

In this case up to K LCs can be multiplexed on each -channel with each LC allocated one subchannel of a -channel. For example, if TDM/TDMA is used, a subchannel would correspond to a time slot. With several -channels allocated to each path, TDM/T-WDMA can be used so that the basic subchannel corresponds to one channel slot.

This procedure is valid for re ning the granularity of any LLN. We shall apply it to two cases of the Petersen network compared in Table 6.6: W = 3 and W = 5.

We determine the traf c-handling capacity of the ne-grained network, assuming that each LC requires one channel slot. Let f E (C, PB ) be the solution of Equation (5.91) for.

Multiwavelength Optical Networks G; that is, f E gives the max Denso QR Bar Code for None imum offered traf c to C channels for a blocking probability not exceeding PB . Then for the cases being considered, the offered traf c per node sustainable by the Petersen network with blocking probability not exceeding PB , using K subchannels per -channel and C -channels per waveband, is G = 9 f E (K C/3, PB ), G = 9 f E (K C, PB ), for (W = 3) for (W = 5). (6.

71). The constants multiplying f E in each equation re ect the traf c splitting factor at each node. The rst argument of f E indicates the number of subchannels available on each path. (For the case W = 3, the argument K C/3 re ects the fact that the C -channels in each waveband must be divided among three sets of connections to satisfy the DCA condition.

For W = 5, we assume full sharing.) Using the values of G from Equation (6.71), the maximum carried traf c sustainable by the Petersen network at a blocking probability PB is S(K , C, PB ) = 10(1 PB )G/K , (6.

Copyright © . All rights reserved.