Performance MultipleTravelTimesConsideration

[Bernd Welter]

Cheers,

these days a partner compared the calculation times of distance matrices within xDIma1 and xDima2. Here's some important feedback about this - related to MultipleTravelTimesConsideration:

• As the usecase is supposed to consider Break- and Restrules (aka WorkingHours) we needed to apply the MultipleTravelTimesConsideration.
• This limits us to a maximum planning horizon of 24hours.

Becoming more flexible requires the engine to compute more data: this has a significant impact on computation time of the Dima. Therefore I'd like to draw your attention to some performance relevant mechanisms:

• Try to apply HighPerformanceRouting if possible (on premise only)
• Try to simplify the calculation by applying one or even several of the following strategies:
  • Reduce the scope of the DIMA dalculation: use geographic restrictions such as "reduction to continent" (allowedCountries={"ZZ_EUR"}), reduction to some countries (allowedCountries={"DE","AT","CH"} or even a small rectrangle (customBounds) - you can apply these in the Creation of the HPR and also in the calculation of the Dima.
  • Be aware that applying too many geographic filters could cause side effects, e.g. the intersection of RECT (area around Berlin) and COUNTRIES (DE,CH,AT) could lead to "waypoints not reachable" which has a negative impact onb performance AND output quality.
  • If your planning horizon's span is less than 24 hours (e.g. 06:00-18:00) you could also reduce the HPR/DIMA's MultipleTravelTimesConsideration interval to this.

Bernd