Back to writing

October 28, 2023 · 5 mins

CASPER and Carbon-Aware Distributed Scheduling

Research from UMass Amherst on scheduling distributed web application workloads while accounting for carbon cost and performance constraints.

Distributed SystemsResearchSchedulingSustainability

Research Context

At the University of Massachusetts Amherst, Basundhara worked with the Lab for Advanced Software Systems on carbon-aware distributed scheduling for web applications.

The research problem was practical: distributed applications need performance, but the infrastructure that powers them also has changing carbon costs across regions and time. Scheduling decisions can account for both.

Project Focus

The CASPER work explored spatial workload scheduling algorithms that factor in carbon cost while still respecting performance constraints for distributed web applications.

The project connected several systems concerns:

  • geographically distributed web workloads
  • performance-sensitive scheduling
  • carbon-aware placement decisions
  • cloud infrastructure efficiency
  • tradeoffs between responsiveness and sustainability

Publication

The work was published on October 28, 2023 in the Proceedings of the 14th International Green and Sustainable Computing Conference, as listed on Google Scholar and Crossref for DOI 10.1145/3634769.3634812.

Public paper: lass.cs.umass.edu/papers/pdf/igsc2023-casper.pdf

Portfolio Relevance

CASPER shows the research side of the same engineering theme that appears in Basundhara's industry work: distributed systems need careful scheduling, reliable communication, and designs that balance real-world constraints instead of optimizing one metric in isolation.