Check out the new USENIX Web site.
2004 USENIX Annual Technical Conference, June 27-July 2, 2004, Boston Marriott Copley Place, Boston, MA
USENIX '03 Home  | USENIX Home  | Events  | Publications  | Membership








Open Sessions











Complete Technical Sessions
By Day: Monday | Tuesday | Wednesday | Thursday | Friday
By Session: General Sessions | FREENIX | SIGs | Guru Is In | WiPs

Locations: See the complete technical sessions.

Monday, June 28
8:45 a.m.–9:00 a.m. Monday

Opening Remarks & Awards
Andrea Arpaci-Dusseau and Remzi Arpaci-Dusseau, University of Wisconsin, Madison

9:00 a.m.–10:00 a.m. Monday
Plenary Session
Open Source and Proprietary Software: A Blending of Cultures
Alan Nugent, VP & CTO, Novell
Listen in MP3 format
Linux/Open Source

As a 20-year provider of proprietary software for the enterprise market, Novell has built products and a culture around proprietary (or closed) software. Within the last 18 months, we have embraced open source development and Linux and have injected them into our corporate DNA. While different, the two approaches are not mutually exclusive. In fact, I would argue embracing open source as a proprietary company is more straightforward than an open source company trying to move "up the stack." In this talk I will examine the myths, challenges, and opportunities for companies attempting to understand the best of both worlds.

Alan F. Nugent serves as chief technology officer of Novell. Prior to Novell, Alan was the Managing Partner, Technology, at Palladian Partners. Mr. Nugent has successfully led many different technology organizations. He serves on the Board of Directors and on the Technical Committee for the Object Management Group and is a widely respected writer and speaker on OT, BPR, and Information Management. He sits on the board of directors of several technology startup companies.

10:00 a.m.–10:30 a.m.   Break  
10:30 a.m.–12:00 p.m. Monday
Instrumentation and Debugging
Session Chair: Val Henson, Sun Microsystems
Making the "Box" Transparent: System Call Performance as a First-Class Result
Yaoping Ruan and Vivek Pai, Princeton University

Dynamic Instrumentation of Production Systems
Bryan M. Cantrill, Michael W. Shapiro, and Adam H. Leventhal, Sun Microsystems

Flashback: A Lightweight Extension for Rollback and Deterministic Replay for Software Debugging
Sudarshan M. Srinivasan, Srikanth Kandula, Christopher R. Andrews, and Yuanyuan Zhou, University of Illinois, Urbana-Champaign


Advanced System Administration SIG
Session: Automating System and Storage Configuration Sysadmin
CHAMPS: A Schedule-Optimized Change Management System
Alex Keller, IBM T.J. Watson Research Center

IBM's CHAnge Management with Planning and Scheduling (CHAMPS) system gleans and exploits detailed structural knowledge about software dependencies in a distributed system. CHAMPS' ability to gather and optimize such information according to administrator-defined objectives, available resources in server pools, and financial constraints enables high-quality optimized solutions for quick and reliable software provisioning tasks, such as installations and updates.

Autonomics in System Configuration
Paul Anderson, University of Edinburgh

A computing installation is created from a collection of bare hardware and a repository of software. These same components may be used to build "fabrics" with very different objectives, from a GRID computing cluster to a network of desktop workstations. We use the term "system configuration" for the task of turning these raw components into an integrated system that satisfies the given objectives. However, we also want to maintain these objectives in the face of changing requirements and external influences. To do this, a system needs to be able to adapt autonomically. This talk discusses some of the problems of specifying and implementing configurations in an autonomic environment.

Embedded Linux
Bart Massey, Portland State University
Linux/Open Source Networking
Bart is the faculty advisor to the Portland State Aerospace Society (PSAS). PSAS has been using Linux as the embedded OS for launch control, flight sequencing, and in-flight telemetry in the extremely unforgiving environment of their 12-foot-high supersonic rocket. Bart has also been involved with the efforts of Keith Packard and Carl Worth in developing the Linux/X infrastructure for embedded devices in conjunction with the project.

The session will be fun-filled and informative for embedded Linux novices and experts alike. Session focus will be on cost-effective, practical Linux solutions for a variety of embedded environments.

12:00 p.m.–1:30 p.m.   Lunch (on your own)  
1:30 p.m.–3:00 p.m. Monday
Swimming in a Sea of Data
Session Chair: Yuanyuan Zhou, UIUC
Email Prioritization: Reducing Delays on Legitimate Mail Caused by Junk Mail
Dan Twining, Matthew M. Williamson, Miranda J. F. Mowbray, and Maher Rahmouni, Hewlett-Packard Labs

Redundancy Elimination Within Large Collections of Files
Purushottam Kulkarni, University of Massachusetts; Fred Douglis, Jason LaVoie, and John M. Tracey, IBM T.J. Watson

Alternatives for Detecting Redundancy in Storage Systems Data
Calicrates Policroniades and Ian Pratt, Cambridge University


Advanced System Administration SIG
Session: System Administration: The Big Picture Sysadmin
The Technical Big Picture
Alva Couch, Tufts University

Network and system administration are still considered bottom-up processes: integrating disparate hardware and software into a usable computing infrastructure, building a pyramid of capabilities and services upon a foundation of reliable core systems. By contrast, the true goal of network and system administration is to support a particular top-down mission or pattern of use, by expressing organizational goals as subgoals in an upside-down pyramid of needs. Looking at the profession from a top-down perspective leads to new metrics for the profession, some surprising and controversial conclusions, and several interesting questions for the future.

The Human Big Picture
Tom Limoncelli, Independent Consultant

Imagine rolling out a security patch, a new application, or a new operating system to 40,000 PCs—it's 65% communication, 25% technical work, and 10% ego management. System administration on a large scale becomes a study of human relationships: managing large teams of people on a project, managing expectations and resources with upper management, and coordinating with a user base. Psychology and public relations become just as important as technical prowess. Why aren't those skills taught to CS majors?

Session TBA
BSD Linux/Open Source
3:00 p.m.–3:30 p.m.   Break  
3:30 p.m.–5:00 p.m. Monday
Network Performance
Session Chair: Carl Staelin, HP Labs
Networking Sysadmin
Monkey See, Monkey Do: A Tool for TCP Tracing and Replaying
Yu-Chung Cheng, University of California, San Diego; Urs Hölzle and Neal Cardwell, Google; Stefan Savage and Geoffrey M. Voelker, University of California, San Diego

A Transport Layer Approach for Improving End-to-End Performance and Robustness Using Redundant Paths
Ming Zhang and Junwen Lai, Princeton University; Arvind Krishnamurthy, Yale University; Larry Peterson and Randolph Wang, Princeton University

Multihoming Performance Benefits: An Experimental Evaluation of Practical Enterprise Strategies
Aditya Akella and Srinivasan Seshan, Carnegie Mellon University; Anees Shaikh, IBM T.J. Watson


Advanced System Administration SIG
Session: Large Storage Sysadmin
Autonomic Policy-based Storage Management
Kaladhar Voruganti, IBM Almaden Research

The cost of managing ever-growing storage is now a dominant IT expense. Storage administration complexity is growing, and the amount of storage a single administrator can manage is not growing as quickly as a site's storage capacity. We present a new approach: autonomic policy-based storage management. This approach dramatically increases administrative leverage by moving storage admin to software that understands high-level policy goals with respect to performance, availability, security, and backup/restore. The talk discusses the architecture, algorithms, and implementation details of a policy-enabled storage-network planner—a key component of the overall policy-based storage management solution.

Experiences with Large Storage Environments
Andrew Hume, AT&T Research

I have worked with various aspects of the recording and billing systems for a large telecommunications company for the last several years. We process feeds from various UNIX, MVS (Cobol and PL/1), and AS/400 systems, to the tune of 4000 files comprising 300–700GB of new data per day. This data streams into several PC clusters, where it is chewed on, compressed, added to databases, and copied onto tape. This is not trivial to do. Sure, we have about 50TB of disk, but that doesn't go as far as it used to. Managing the space is harder than it ought to be; annually managing 2–3 million files and their life cycles is awkward, as is controlling the processing of these files. This talk will discuss these and other challenges in both hardware and operating system choices and will discuss solutions we've found.

?Need help? Use our Contacts page.

Last changed: 19 July 2004 aw