Performance
Tractable problems
For problems such as distance matrix generation, we use sophisticated indexing systems which help achieve great performance on producing a large number of pairwise, road network distances between points.
Intractable problems
Many of the problems our APIs handle are NP-Complete which means that for bigger problems, we’re sometimes not able to prove that we have an optimal solution to the problem provided. That said, we have a pool of heuristics that we use to find good solutions quickly in these cases. Often, these heuristics are being driven by information about where the optimal could possibly still be - this helps keep the search effective in the time constraints.
For each model in the gallery you’ll find guidelines for typical computation times. If you find that your particular model is falling far outside these bounds, feel free to drop us a note and we’ll investigate the performance of your model.
API Calls
If you’re using the protobuf interface you would have already experienced really fast data transfer times for post and get requests. This is mostly driven by the compressed packet size being used by protobuf, but also the deserialisation requires minimal parsing because of the compliance to the specified schema.