### Table 2: Parameters of the 39 services using the NSFNET backbone. We assume a source node considers only the set of minimum-hop and minimum-hop+1 paths for connection routing. A path from the set is selected probabilistically according to path weights as explained in Section 4. The rst selection algorithm, referred to as SEL.HOP, de nes the path weight as 1=Hp. The second selection algorithm, referred to as SEL.UTIL, de nes the path weight as (1 ? Up), where Up is the maximum link utilization of the links on path p. The third selection

1994

"... In PAGE 17: ...tep equals 0.1. The routing update period equals 5. We consider 39 services using the NSFNET backbone, with parameters as shown in Table2 . Services with the same tra c and end-to-end QoS parameters, but with di erent source/destination pairs, are grouped in the same row.... ..."

Cited by 4

### Table 1 such that all values are non-negative. From the set of feasible paths we select a path probabilistically according to the resulting distribution. In general, for S classes of VC requests, if Rk is the smallest route set to which a path p belongs, then the weight given to select p, W(p; k), is given by:

1998

"... In PAGE 4: ...For a request distribution with half the granularity{i.e. a uniform distribution over the [0 ? 0:5] interval|all VC requests will be possible to honor when the system utilization is 50%, and one fth when the system utilization is 90%. Smallest route set Weight of choosing the path R1 d1 + d2 + d3 + d4 R2 d2 + d3 + d4 R3 d3 + d4 R4 d4 Table1 . Weight assigned to various routes.... ..."

Cited by 12

### Table 1 such that all values are non-negative. From the set of feasible paths we select a path probabilistically according to the resulting distribution. In general, for S classes of VC requests, if Rk is the smallest route set to which a path p belongs, then the weight given to select p, W(p; k), is given by:

1998

"... In PAGE 4: ...For a request distribution with half the granularity{i.e. a uniform distribution over the [0 ? 0:5] interval|all VC requests will be possible to honor when the system utilization is 50%, and one fth when the system utilization is 90%. Smallest route set Weight of choosing the path R1 d1 + d2 + d3 + d4 R2 d2 + d3 + d4 R3 d3 + d4 R4 d4 Table1 . Weight assigned to various routes.... ..."

Cited by 12

### Table 1: Results obtained using the uniform random selection method for the 20 node network. The number of paths indicates the number in each route for a node pair. The initial blocking is the result of initialisation using our heuristic method. The nal blocking is the result produced by SA. As expected, the call blocking decreases monotonically when the number of avail- able indirect paths increases. This also implies that better utilisation of link capacities can be achieved with more indirect paths. 3.2 Probabilistic Selection Method The uniform random selection method is simple, but it ignores the current status of a network, i.e., the path capacity and call request. The probabilistic selection method attempts to make use of such information in order to generate more promising next

"... In PAGE 10: ... The nal blocking is the result produced by SA. Comparing Table 2 with Table1 , it is clear that the probabilistic selection method has produced better results than the uniform random selection method does. We have also conducted simulation on another probabilistic selection method which always as- signs 0:8 probability to the direct path and assigns the remaining 0:2 probability... ..."

### Table 3: Results obtained using the probabilistic selection method which encourages utilisation of direct paths. The number of paths indicates the number in each route for a node pair. The initial blocking is the result of initialisation using our heuristic method. The nal blocking is the result produced by SA. Somewhat surprisingly, the encouragement of using direct paths has led to signif- icant better routing schemes in our experiments. This is probably because a direct path uses only one link while an indirect path occupies two.

### Table 4: Simulation results on an existing 7-node network with di erent call requests. The total link capacity of the network is 521. Blocking 1, 2 and 3 list results generated, respectively, by SA with uniform random selection method, SA with probabilistic selection method and SA with bias towards direct paths. The results were averaged over ten runs.

### Table 2 : Probabilistic paths through each task node in the Information Search

"... In PAGE 5: ... Table 2 : Probabilistic paths through each task node in the Information Search While we can see in Table2 that the probability of a scan being found in the radiology archive is high, the small likelihood (.05%) of the search having to move onto the rescan node carries a high cost (20 days: Table 1 ) Examining the various possible paths in Figure 1 and accounting for the probabilities of these being followed, we found a weighted average of nearly 6.... ..."

Cited by 1

### Table 8: Probabilistic model: Terminology

"... In PAGE 38: ... Notation P denote the path being currently considered. Table8 summarizes the notation. 1 1 1 0.... ..."

### TABLE II Probabilistic prediction of symbols on path until next update

### Table 7.6: Expected Travel Time and Population on Paths 1 to 6 for Time Interval 1 Path 1 Path 2 Path 3 Path 4 Path 5* Path 6*

1998

Cited by 1