Cornell-UC Berkeley
Research on Resource Disaggregation

About

Existing datacenters are built using servers, each of which tightly integrates a small amount of the various resources needed for a computing task (CPU, memory, storage). While such server-centric architectures have been the mainstay for decades, recent efforts suggest a forthcoming paradigm shift towards a disaggregated datacenter, where each resource type is built as a standalone resource "blade" and a network fabric interconnects these resource blades. Examples of this include Facebook Disaggregated Rack, HP "The Machine", Intel Rack Scale Architecture, SeaMicro as well as prototypes from the computer architecture community.

Projects

Our research targets every aspect of resource disaggregation - feasibility requirements, network fabric design, storage fabric design and compute fabric design.

Network Requirements for Resource Disaggregation

The focus of this project is to derive the minimum latency and bandwidth requirements that the network in disaggregated datacenters must provide to avoid degrading application-level performance and explore the feasibility of meeting these requirements with existing system designs and commodity networking technology.

Network Fabric for Resource Disaggregation

1. Shoal

This project focusses on designing predictable low latency high bandwidth network fabric for high-density disaggregated racks within datacenters.

2. vRack

This project focusses on building flexible racks within datacenters, where groups of physically apart servers can be put into virtual racks to enhance their performance.

Compute Fabric for Resource Disaggregation

1. Monotasks

This project focusses on how to reason about the performance of today's systems by making it easy to understand where bottlenecks lie and the performance implications of various system changes.

People

Our research team combines expertise in systems, networking, and computer architecture.

Faculty

Rachit Agarwal

Cornell

Scott Shenker

UC Berkeley

Sylvia Ratnasamy

UC Berkeley

Christina Delimitrou

Cornell

Hakim Weatherspoon

Cornell

Students & Researchers

Vishal Shrivastav

Cornell

Peter Xiang Gao

UC Berkeley

Kay Ousterhout

UC Berkeley

Alana Marzoev

Cornell

Lloyd Brown

Cornell

Justin Miron

Cornell

Saksham Agarwal

Cornell

Publications

Performance clarity as a first-class design principle
Kay Ousterhout, Christopher Canel, Max Wolffe, Sylvia Ratnasamy, Scott Shenker.
HotOS 2017
Network Requirements for Resource Disaggregation
Peter X. Gao, Akshay Narayan, Sagar Karandikar, Joao Carreira, Sangjin Han, Rachit Agarwal, Sylvia Ratnasamy, Scott Shenker.
OSDI 2016

Industry

Our research has strong ties to industry.

Companies

People

Amin Vahdat

Google

Hitesh Ballani

Microsoft Research

Paolo Costa

Microsoft Research