The Operating Systems group at Cornell examines the design and implementation of the fundamental software systems which comprise our computing infrastructure. Our interests span from secure OS's and languages, to data-center scale management and consistency systems, to global routing algorithms. Increasingly, trustworthiness, broadly construed, permeates our systems work. Security research across the department, including systems, is described here .
Some of our ongiong projects are described below: The Operating Systems group at Cornell examines the design and implementation of the fundamental software systems which comprise our computing infrastructure. Our interests span from secure OS's and languages, to data-center scale management and consistency systems, to global routing algorithms. Increasingly, trustworthiness, broadly construed, permeates our systems work. Security research across the department, including systems, is described here . Some of our ongoing projects are described below: The QuickSilver project is developing a new platform to support scalable, secure, self-repairing applications. For example, our new Live Objects platform makes it easy to build and share distributed applications that incorporate live data. The project is really a "catch all" for a whole collection of subprojects, many of which are quite substantial. Participants include Ken Birman , Robbert van Renesse and Hakim Weatherspoon.
Here are a few of the main activities within Quicksilver:
As mentioned, we've created a new platform that we call "live distributed objects" (http://liveobjects.cs.cornell.edu). This support a drag-and-drop paradigm for building distributed systems using a mashup technology very similar to the ones used in the web, for example with Google's mashup maps.
Quicksilver Scalable Multicast and its successor, QS/2, are reliable group communication platforms designed to support data replication (for example, within our live objects environment).
Ricochet is a new communication protocol we've developed for use in large data centers that need very high-speed event notification in small groups of replicated servers.
Maelstrom is a technology for improving the quality of communication between data centers connected by wide area links that run at very high speeds, but have loss problems.
The Smoke-and-Mirrors and KyotoFS are two projects that use Maelstrom in support of file systems mirrored between datacenters separated by fast but high latency WAN links, or that need to minimize energy consumption by cooperating to spin down disks without reducing data availability. Nysiad is a new technique for transforming a scalable distributed system or network protocol tolerant only of crash failures into one that tolerates arbitrary failures, including such failures as freeloading and malicious attacks. We are investigating the building blocks of consensus protocols and use these building blocks to assemble a skeleton that can be configured to produce, among others, three well-known consensus protocols: Paxos, Chandra-Toueg, and Ben-Or. Fireflies is a scalable protocol for supporting intrusion-tolerant network overlays. Fireflies provides correct nodes with a reasonably current view of which nodes are live, as well as a pseudo-random mesh for communication. Paul Francis ' group works on a variety of scaling, management, and security issues that plague enterprise and ISP networks. Network management is arguably the most expensive and least robust aspect of networking. The CONMan project is looking at fundamental new architectures for network management. Scaling global IP routing has gone unsolved for two decades. We think we've nailed it with an approach we call Virtual Aggregation, which lets ISP autonomously scale routing using existing equipment and protocols. The ShutUp Protocol exploits virtual machine technology to create a pure end-to-end solution to unwanted Internet traffic, including DoS and flash worms. Priority-Layered Transport (TLP) is a transport protocol for long-fat pipes that matches the performance of XCP but without requiring changes to routers. Gun Sirer' s group is working a range of peer-to-peer and security problems. Beehive is a high-performance, self-organizing, scalable, peer-to-peer distributed hash table (DHT). Unlike other peer-to-peer systems, whose performance scales with O(log N), Beehive achieves O(1) lookup performance for certain, commonly-encountered query distributions. The Herbivore project is a distributed, peer-to-peer anonymous communication system, providing private file sharing and messaging over the Internet. It lets people anonymously publish and retrieve documents against even the most resourceful adversary. The MagnetOS project is building a new operating system for ad hoc and sensor networks. MagnetOS is a distributed operating system for ad-hoc and sensor networks whose goal is to enable power-aware, adaptive, and easy-to-develop ad-hoc networking applications. Graduate students who join our department will have a chance to shape the future of our computing infrastructure. Researchers Ken Birman
Distributed computing, fault-tolerant network systems, distributed systems security, large-scale network applications. Paul Francis
Network management, network routing, high-performance transport, unwanted network traffic. Andrew Myers
Programming languages, security, mobile code, persistent and distributed objects. Robbert van Renesse
Distributed computing, peer-to-peer networking, scalability, fault tolerance, adaptive networking. Fred B. Schneider
Distributed systems security and fault-tolerance, mobile code, concurrent programming, secure OS. Emin Gun Sirer
Operating system support for ad hoc networks, peer-to-peer systems, self-organizing overlays, networked services and extensible systems, secure OS. | 
