Data indexing - the way forward

Send to friend

Retired university lecturer Jim Williamson has developed a new approach to indexing of data, with positive implications for the warehousing and logistics industry.

Currently index files are used to provide very fast access to data related to specific keys or to put data records into a particular sequence. Their use with keys underpins the operation of relational databases and they are used widely within logistics applications. They can be used to support the concept of 'nearest' and return data from records that does not match the key but is the closest to the value of the key. The key can consist of more than one field but these are treated in a hierarchy.

The index files do not allow for a location search by latitude and longitude with equal significance being placed on both values. This problem has been solved by the geospatial index, which can return a record that has data describing something closest in distance to the sought location. However, there is no available index to retrieve data by location and date. This concept has an extra problem in that time and distance are not measured in the same units, and are not capable of being compared. If we are looking for proximity, which is closest – '2 days and 30 kilometres, or 3 days and 20 kilometres'? This can be solved within any specific context with such pragmatism as 'one day equals 40 kilometres'.

The existence of such an index can be used to improve the effectiveness of such things as event listings or news reporting, where the results of a search are immediately presented to a user. There are possible uses for this index in systems where the results of a search are processed by a program rather than being shown to a user. If an organisation has many members of staff working at different locations and there are new jobs appearing in real-time then automation can provide the dynamic scheduling of staff. This style of solution has an extra advantage when it is not known how long tasks will take.

One logistics problem that computers have successfully solved is vehicle routing, but the solution pre-supposes that all the destinations are known in advance. There may be a need for vehicles on the road to also visit destinations that were not known about when the route was scheduled. If there is a need to select a vehicle to make an extra stop – perhaps to pick up some parcel or goods –then the index could be used to select the vehicle that will be closest to the given point at some specific time.

This switch from pre-processing a batch of jobs to interactively processing a single task is a change in programming style that is well-understood. Many transport tasks are performed with increased efficiency in batch mode and the world of logistics reflects the need for real-world goods to be batched for efficient delivery. Just-in-Time manufacture does not require individual wing-mirrors to be delivered at the speed of the car production line. However, dealing with tasks in the 'here and now' requires immediate decisions – whether by humans or by software.

Another potential application would be the allocation of tasks for the 'return journey' of vehicles. If the data concerning 'transport needed' were available based on pick-up location and destination location, then the computer could quickly select suitable tasks. The general adoption of this approach could be that software 'logistics agents' would select both transporters and warehouses for tasks, but this requires the solution of trust issues that are more complex than programming problems.

Another style of solution could be the provision of shared door-to-door transport as a general form of public transport. Door-to-door transport is provided by specialist operators for people who have problems using public transport and also in rural areas. These systems do not involve large numbers of vehicles and so the computer tends to be used to support and record decision making. A general system could be provided with would-be travellers indicating their transport needs via an app and the computer system automatically selecting a vehicle for shared use. This would require a search for the vehicle that was closest to the pick-up at a specific time and the destination. Once again, there are many non-technical issues for such solutions including the role of competition and legal aspects.

There are no doubt many other problems that become soluble in a new manner with the use of multi-dimensional indices. This article has been written to promote thinking about new approaches by system designers.

About the author

Jim Williamson has presented conference papers on a range of transport topics as part of his academic career. As the owner of Blue Collar Software, he designed the first program in Western Europe to allow users to interactively schedule journeys within a Dial-a-Ride operation. This software ran on a network of PC computers more than twenty-five years ago.

For further information please contact

Comments (0)

Add a Comment

This thread has been closed from taking new comments.