Analysis of Alibaba Traces

Abstract

Quals talk:

Microservices are the de facto approach to build distributed systems in industry today. Major organizations including Netflix, Facebook, Google and Microsoft are built around microservices for the benefits of agile development, elastic scaling, and fine-grained fault-tolerance. Microservices are loosely-coupled services that interact via languages-agnostic protocols. Beyond such a basic design principle, many microservice characteristics are unclear, yet they are needed for research. In particular, topological characteristics, drawn from request workflows in microservice applications, have been heavily used in recent research to study autoscaling and resource management. To find such characteristics, recent research projects have been primarily relying on open- source testbed applications. However, our recent paper identified notable mismatches between these testbeds and industrial perceptions of microservice applications. To this day, the only way to gain topological characteristics of industrial, large-scale microservices is to use the recently-released Alibaba microservice dataset.

In this talk, I will present a systematic approach to reconstruct workflow traces from Alibaba microservice dataset, along with their topological characteristics.

Research area: Systems

Please join meeting in Cummings 302, or via Zoom.

Join Zoom meeting: https://tufts.zoom.us/j/9850318491

Passcode: please see colloquium email

Dial-in not an option for this event.