PIPages
Labs

Minlan Yu

Harvard University, MA·Computer Science

Cloud data centers run social media, video streaming and artificial intelligence, and they are struggling to keep up with demand. Normally, the servers do all the computing while network switches, the devices that pass data packets between machines, only route traffic.

Fields

Computer Science

Engineering

Topics

computer networking

in-network computing

programmable data plane

Grants
2 known grants, through Sep 2027

What they work on

  1. Design a stateful network switch architecture that can manage, compute on and share application state inside the network, enabling in-network computing for distributed and high-performance computing applications.
  2. Create a network-wide abstraction and runtime that lets programmers operate many stateful switches together across a data center, and evaluate architectural trade-offs with a suite of representative applications.
  3. Design approximation and randomized algorithms for programmable data plane switches that improve network telemetry and other distributed network applications while fitting within tight memory and computing limits.
  4. Build libraries of switch-level utilities and data structures for approximate and randomized algorithms, and develop theoretical frameworks to guide algorithm design for current and future switch architectures.

More about the research

Newer programmable switches can run custom code as data flows through them, opening a chance to do useful work inside the network itself. But these switches have little memory and limited computing power. Both projects ask how to make network hardware do more within those tight limits.

Why it matters

Data centers and the wider Internet underpin health, transportation, energy, education and business. Switches that compute and track state could speed up large-scale and high-performance computing used by researchers, government and companies. Resource-efficient approximate and randomized algorithms could make networks more efficient, robust and secure, with better monitoring and troubleshooting, and shared libraries could help others adopt these methods.

Known grants

From NIH and NSF research grants. Labs often have other funding we can't see.

Collaborative Research: CNS Core: Medium: A Stateful Switch Architecture for In-Network Compute

NSF · Oct 2022 – Sep 2027

CNS Core: Medium: Approximation and Randomization in the Programmable Data Plane

NSF · Jul 2021 – Jun 2027

Closest labs

Oyekunle Olukotun

Stanford University, CA·Computer Science

Design a stateful network switch architecture that can store, compute on and share application data inside the network, beyond basic packet routing.

Reza Rejaie

University of Oregon, OR·Computer Science

Design a multi-objective, approximation-based scheme that schedules monitoring queries on a single multi-stage programmable switch, balancing accuracy, latency and reporting load.

Arvind Krishnamurthy

University of Washington, WA·Computer Science

Design NetFusion, a rack-level pool of programmable network interface cards (SmartNICs) that perform packet and network-function processing for many servers, with safe and fair sharing among applications.

The school

Harvard University: 42 labs on PI Pages, $1.5B a year on research, 712 research doctorates a year.

More about Harvard University

Reach out

A start for your email. Fill in the brackets.