University of California at Berkeley, Computer Science Division
Ph.D. (major: Operating Systems,
minor: Management of Technology), 2001.
M.S., 1998.
University of California at San Diego, Linguistics Department
B.A. Summa Cum Laude and With Highest Distinction
(Concentration in Language and Computers), 1991
Design and implementation of highly available, high-performance, scalable systems.
Strong background in Unix internals, networking, parallel and distributed systems.
Proficient in C, C++, Lisp, and Perl. Familiar with C# and Java. Unix/Linux and Windows environments.
Senior Program Manager, Microsoft Search, Relevance (2/2009 - present)
Work on the alterations part of the Bing search relevance
pipeline. Responsible for analysis and improvements to the speller
service, including analyzing large data sets, evaluating the
performance of each speller component, experimenting with new training
techniques, collaborating with international teams, collecting and
processing human judgment data, and maintaining the live service.
Senior Program Manager, Microsoft Search Infrastructure (1/2006-1/2009)
Worked on the Cosmos distributed store and computing platform. I was
the sole PM for the majority of the project while we went from a 10
machine test system to a 30,000+ machine base across 3 data centers,
reliably storing several PB of data and processing thousands of
parallel jobs per week. I made direct contributions to customer
requirements understanding, design of the store, analysis of
distributed programs, and automated monitoring tools. The code base
from this project was used for Microsoft’s Azure cloud store.
Senior Software Engineer, Scale Eight, Incorporated, 2001-present
Principal designer of a commercially-deployed distributed file
system. Designed and implemented highly scalable, fault-tolerant,
distributed services for caching, metadata management, locking, and
cluster membership. Developed tools for characterizing performance and
detecting bottlenecks.
Chief Architect, Market Engine Corporation, 1999-2001
Designed a distributed infrastructure for a cluster of workstations,
including fault-tolerance and load-balancing mechanisms. Designed and
constructed the Java-based distributed infrastructure, including
messaging and fault-tolerance software. Served as liaison between
engineering team and customers, including tasks such as presenting
technical overviews to customers and answering pre-sales questions.
Graduate Student Researcher, Computer Science Division, 1994-2001
University of California at Berkeley, Network of Workstations (NOW) project
Designed the xFS serverless, distributed file system and implemented
the storage and networking components. Wrote a successful NSF grant
proposal to create a large-scale filesystem tracing
project. Instrumented Unix kernels to collect long-term file system
traces, created a framework for characterizing Unix and NT application
workloads, and compared long-term performance of conventional and
novel file system designs.
Teaching Consultant, National Technical University, 1999-2000
Instructor for graduate class in computer architecture for a distance
learning course attended remotely by professional engineers. Wrote
homework and exams, graded course work and projects, mentored
engineers in course work.
Graduate Student Instructor, Computer Science Division, 1993-1994
University of California at Berkeley, Operating Systems Course
Lectured on operating system concepts, managed the system simulation
project, and supervised graders for a class of 90-130 undergraduates.
System Administrator, Molecular and Cell Biology, Summer 1993
University of California at Berkeley
Solely responsible for the department's computer systems.
Provided all hardware, software, and administrative support for a
network of SGI workstations running Irix and DECStations running VMS.
Intern, Lawrence Berkeley Laboratory, Summer 1992
Imaging Technologies Group
Developed programs to automatically perform clustering and
segmentation on 2D MRI data as a step in building a manipulable, 3D
model. Analyzed the effectiveness of various statistical strategies
against noise.
Research Fellow, San Diego Supercomputer Center, Summer 1991
With a research fellowship awarded by the National Science
Foundation, designed and implemented a program to parse phonetic input
into a grammar using statistical techniques.
Engineer, Electrical and Computer Engineering, 1990-1991
University of California, San Diego
Implemented an expert shell to provide a
knowledge database and intelligent interface for various CAD tools.
``Long-term File System Performance,'' D. Roselli, Dissertation, May 2001.
``A Comparison of File Systems Workloads,'' D. Roselli, J. Lorch, and T. Anderson, Proceedings of the Annual USENIX Technical Conference, June 2000.
``Self-Similarity in File Systems,'' S. Gribble, G. Manku, D. Roselli, E. Brewer, T. Gibson, and E. Miller, Proceedings of ACM SIGMETRICS, June 1998.
``Improving the Performance of Log-Structured File Systems With Adaptive Methods,'' J. Matthews, D. Roselli, A. Costello, R. Wang, and T. Anderson, Proceedings of the Sixteenth ACM Symposium on Operating System Principles, Saint Malo, France, October 1997.
``Long-term File System Tracing,'' D. Roselli, Master's Project Report, University of California at Berkeley, May 1997.
``Serverless Network File Systems,'' T. Anderson, M. Dahlin, J. Neefe, D. Patterson, D. Roselli, and R. Wang, ACM Transactions on Computer Systems, February 1996, and ACM Proc. 15th Symposium on Operating Systems Principles, December 1995. Award paper.
``Pinta: A System for Visualization of Anatomical Structures,'' B. Parvin, W. Johnston and D. Roselli, IEEE Conf. on Computer Vision and Pattern Recognition, June 1993.
"A Distributed Network Attached Storage System." A method for implementing a scalable NAS system using a distributed file system with a distributed volume manager, Scale 8, Inc., February 2003.
"Methods and Apparatus for Load Balancing Storage Nodes In A Distributed Network Attached Storage System." A method for exporting a single system image for a scalable NAS system, Scale 8, Inc., February 2003.
"Methods and Apparatus For Load Balancing Storage Nodes In A Distributed Storage Area Network System." A method for exporting a single system image for a scalable NAS system using a SAN for storage, Scale 8, Inc., February 2003.
"Fault-Tolerance Framework for an Extendable Computer Architecture." A hierarchical method of load balancing and fault tolerance in a distribute work system. Market Engine Corporation, March 2001.