Investigate synchronous bulk geocoding

The objective of this task is to investigate the feasibility of synchronous bulk geocoding compared to asynchronous batch geocoding. In synchronized bulk geocoding, the online geocoder is extended with a new resource named:

addresses/bulk

This resource implements a POST request that takes a short list of addresses, geocodes them, and returns the list of geocoded results in the response. This is a stateless request; no jobs are created or managed. Handling of very large address lists is handled through an bulk address list submitter script that is written in python and reads in the source address list, issues a series of addresses/bulk requests, and outputs the geocoded results to a single file.

Performance tests should be conducted with various settings of address list size and time budget. For example, set global config parms maxAddressListSize = 100 and timeBudget=20 (seconds) and run address lists of sizes ranging from 1000 to 10 000 000. In a single request, the bulk geocoder would only do as many addresses as can be completed within the timeBudget and to a maximum of 100 addresses. The timebudget is critical as there are various server timeouts in the entire round-trip processing that must be honoured or the hard work of the bulk geocoder would be lost to oblivion.

The objective of this work is not to have delivery of production code, just determination of the feasibility of bulk geocoding. The bulk geocoder need only support CSV and Geographics projection (4326).

Deliverables should include address/bulk resource, bulk address list submitter script, determination of feasibility of bulk geocoding, comparative timings between bulk and batch geocoding, and suggestions for implementing production versions of both the bulk geocoder and the bulk address list submitter.