Paya: Availability at the Edge

While edge computing is touted as the next big evolution in cloud computing, little is said about edge availability. Customers use the cloud for the operation of their businesses and they expect the cloud to be available 24x7x365. Should they not expect the same from their edge servers? However, in a single edge deployment, many diverse applications share the same edge cluster while having different requirements on availability. For example, a building heating application that handles AC/HVAC does not care if it loses a few temperature readings, while a manufacturing monitoring application cares for every reading. Also, applications might have a state that they should not lose, e.g., a manufacturing application counting the number of items produced. Deciding how to make an application resilient considering its unique requirements is a non-trivial task. Often, developers have to build tailored solutions for each application and each deployment structure.

Paya is an edge-tailored solution that ensures the high-availability is met for all applications based on the application’s specific need while imposing minimal overhead on the application developer. Rather than the traditional notion of “9’s of availability”, Paya introduces a more expressive availability service level agreement (SLA) that captures the intricacies of each application. Internally, Paya automatically decides the details of the deployment to ensure the SLA: how many replicas, where to place replicas, and what replication mechanism to use.