Robert Cordingly
PhD Student, Computer Science & Systems · University of Washington Tacoma
About
I'm a PhD student at the University of Washington Tacoma, advised by Prof. Wes Lloyd. My research builds tools and models that make serverless computing platforms more observable, predictable, and accessible. My goal is to empower developers to easily understand cost, performance, and infrastructure trade-offs of serverless platforms.
Across roughly twenty publications I've worked on serverless performance modeling, Function-as-a-Service observability, multi-cloud "sky" computing, and investigating the carbon implications of global workload distribution. Alongside research, my work focuses on creating open source tools that aid the research community in observing serverless platforms, with the Serverless Application Analytics Framework (SAAF) and the FaaS Experiment Toolkit (FaaSET) being the most notable tools. I am also a Predoctoral Instructor in the Graduate Certificate in Software Development Engineering program, where I've taught TCSS 501 (Algorithms) and TCSS 505 (Systems).
I completed my M.S. (2020) and B.S. (2018, Cum Laude) in Computer Science & Systems at UW Tacoma.
General Exam Abstract
Serverless Function-as-a-Service (FaaS) platforms allow developers to deploy applications as functions that are automatically managed and scaled, and that are billed only for the memory and execution time each request consumes. FaaS platforms can host small, infrequently used applications for practically no cost while also scaling to thousands of concurrent instances for massively parallel workloads. However, developers must still choose which cloud provider, which regions, and which function configurations (e.g., memory reservation size) to deploy to, and each choice has performance, cost, and carbon consequences that are difficult to predict. FaaS platforms obfuscate their underlying infrastructure and bill at millisecond granularity, leaving developers to make ad hoc decisions and often to learn the true cost of a workload only after receiving the bill.
To address these challenges, we propose serverless sky computing: a paradigm that aggregates serverless resources across cloud providers, cloud regions, and function configurations, and routes each request to the deployment that best satisfies its objectives for response time, hosting cost, and carbon footprint. FaaS platforms charge nothing for idle deployments, so an application can be replicated across hundreds of regions and configurations at no additional cost, making the routing decision the primary lever for optimization. Our completed work built the sky mesh, a deployment of over 1,400 functions spanning AWS Lambda, Google Cloud, IBM Code Engine, and DigitalOcean Functions, and demonstrated that aggregation-aware routing reduces the carbon intensity of a globally distributed application by up to 99.8%, network latency by 65%, and hosting costs by 58%, with an additional 18.2% cost savings from exploiting hidden hardware heterogeneity across regions. This proposal presents our research questions on (1) serverless resource aggregation, (2) aggregation strategy trade-offs, (3) platform-neutral abstractions, (4) autonomously learned routing policies, and (5) extending the sky to agentic AI workloads.
Research Interests
Serverless computing Cloud performance modeling FaaS observability Sky / multi-cloud computing Federated computing Green & sustainable computing Developer tooling for distributed systems
Open-Source Projects
SAAF — Serverless Application Analytics Framework
A Function-as-a-Service profiling framework for Node.js, Python, and Java functions deployed to AWS, Google Cloud, IBM Cloud, and Azure. Captures fine-grained runtime metrics that enable platform performance evaluation and cost / performance modeling.
FaaSET — Function-as-a-Service Experiment Toolkit
A Python framework that integrates with Jupyter notebooks to streamline serverless development — deployment, testing, benchmarking, and analysis from a single workspace. Used in tutorials and demos.
FaaS Runner
A client application that creates and automates reproducible experiments across Function-as-a-Service cloud platforms. Recently extended into Distributed FaaSRunner for multi-node, multi-threaded endpoint testing (IC2E 2025).
EasyRL Cloud
Cloud deployment for EasyRL — a framework for training, testing, and experimenting with reinforcement learning models.
Publications
Bold author names indicate me; First author tags mark first-author works. Each entry links to the PDF, Slides, conference video, and indexed venue (ACM DL / IEEE Xplore) where available.
Master's Thesis
Teaching
Honors & Awards
Professional Service
Reviewer: 32nd Internation European Conference on Parallel and Distributed Computing (Euro-Par 2026) · Future Generation Computer Systems — The International Journal of eScience (2025) · 15th ACM/SPEC International Conference on Performance Engineering (ICPE 2024) · 12th IEEE International Conference on Cloud Engineering (IC2E 2024) · 11th IEEE International Conference on Cloud Engineering (IC2E 2023).
Education
Curriculum Vitae
A complete and continually updated CV is available as a downloadable PDF.