A BSR application only polls the switches that are likely to provide the best service at any given time. If a remote switch is polled and returns an adjusted EWT greater than that of the current best resource, polling of the remote switch is suppressed for a period of time proportional to the difference between the two adjusted EWT values. In other words, polling of a given location is suppressed whenever the adjusted EWT returned by that location is subsequently replaced by a better adjusted EWT from another resource. The consider step for this location is skipped during the period and vector processing continues at the next step. When the suppression period is over, the consider step once again polls the location. If the location returns the best adjusted EWT, the next call processed by the vector also causes the location to be polled. If the location is not the best, polling is temporarily suppressed.
If no calls are in queue at the remote location an agent can become available at any moment, and BSR, therefore, does not suppress polling for longer than 5 seconds in such situations. BSR does not suppress polling of any remote location for more than 60 seconds, regardless of the differences between adjusted EWT returned by different switches.
Other conditions can also suppress status polls to a location:
Resource exhaustion, that is, no trunks available, queue full.
Administration errors, that is, badly written vectors, or no application plan.
This feature significantly reduces the average number of status polls placed per call. The greater the call volume, the greater the percentage reduction. Let’s take another look at the vector in screen 2.
Let us say the network is operating in a balanced state. EWTs are 30 seconds at all locations and a call arrives every 3 seconds at each site. Adjusted EWTs are 30 seconds at the origin switch and 60 seconds for each remote switch. After each status poll under the conditions, polling is suppressed for 30 seconds. Each remote location is polled, therefore, by every 10th call. On average, this means that each call polls any one location 0.1 times. Since there are four consider steps, each call makes 0.4 polls. Remembering the 1-second polling response time given at the beginning of the example, the average time added to call processing for each call is 0.4 seconds.
The 1st-found available agent strategy, discussed in BSR can cut average polling times further. With the 1st-found strategy, BSR skips all subsequent consider steps in a series if a resource with an available agent is found and deliver the call to that resource.
