ColdNotify: A Notification Service For A Distributed Arctic Observatory

Abstract

One of the key challenges in the Distributed Arctic Observatory (DAO) project is designing infrastructure to reliably interact with remote, configurable observation units that capture and provide observation data from challenging environments. DAO’s infrastructure is a work in progress and researching alternative strategies for interacting with observation units is necessary to gain experience and knowledge about limitations and requirements.

In client-server models, a common approach to keeping clients up to date is continuous polling, however, this may cause unnecessary stress and bandwidth as DAO scales to hundreds or thousands of observation units. Another approach to this is server-initiated publishing methods, where back-end applications provide new data to observation units. This, however, requires per-application implementations that have to keep track of which observation unit has received what, handle unreachable units and potential state loss.

This thesis has explored how notification services can help back-end application reliably interact with observation units in future deployments, to keep them up to date with configurations, perform remote operations or gather data, as DAO scales.

ColdNotify is an application-neutral notification service, based on Thialfi by Google, that aims to reliably deliver notifications to observation units, despite unreliable connectivity and state loss.