Flight Optmisation using in-memory hashing

Abstract

Current travel booking systems take an appreciable amount of time (often over 30 seconds) to process even a relatively straightforward query. Producing results faster will give a competitive advantage in the travel market. Traditionally, data is placed in storage and then accessed and acted upon in the computer's memory which is a reason for the reduction in speed. The emergence of multi-core processors and availability of large amounts of main memory at low cost allows even large data sets such as airline flights and fares to be stored in-memory and processed using many-core processors. In this work we use in-memory technology to develop a fast hash-based algorithm to search for flights and fares in air travel planning. The objective of this project is to select a reasonable-sized set of priced itinerary solutions from the huge number of all possible solutions. Generated priced solutions must satisfy a set of constraints (such as maximum number of connecting points should be equal or less than three, departure station, arrival station, departure date, one way or round trip, maximum price, maximum duration, etc.). A graphical user interface can be created to manipulate this set. Availability checking and flight booking were considered beyond the scope of this project. We show that air travel planning can be implemented using in-memory hashed lookups, eliminating the need for database lookups, searches or graph operations. With the use of in-memory technology, hashing and lookups, priced solutions for a query are generated on average within 40-50 milliseconds. Memory is used efficiently and the total memory consumption for the full data set is approximately1GB which is easily available on a standard PC. This algorithm can also be extended to take advantage of many-core processors such as