CS262a:
Advanced Topics in Computer Systems

Mon/Wed 1-2:30 in 306 Soda
CCN: `26800`

Prof. Eric Brewer, Fall 2008

Papers due anytime on Thursday December 18th.

CS262a is the first semester of a year-long sequence on computer systems research, including operating systems, database systems, and Internet infrastructure systems. The goal of the course is to cover a broad array of research topics in computer systems, and to engage you in top-flight systems research. The first semester is devoted to basic thematic issues and underlying techniques in computer systems, while the second semester goes deeper into topics related to scalable, parallel and distributed systems. The class is based on a discussion of important research papers and a research project.

Prerequisites: The historical prerequisite was to pass an entrance exam in class, which covered undergraduate operating systems material (similar to UCB's CS162). However, this year we are trying it without an exam. However, if you have not already taken a decent undergrad OS class, you should talk with me before taking this class. The exam also had the benefit of "paging in" the undergrad material, which may have been its primary value (since the pass rate was high).

Exams, Papers, etc.

The main work of this class is to read frequently and deeply, while working toward a group research project of publishable quality. Each student will be individually responsible for writing up a short summary of every paper. There will be one midterm exam (albeit near the end), and no final exam. Research projects are a critical aspect of the course. Your goal is to do some quality systems research; that is, to add to our understanding of how to build systems. Research projects must be written up in a term paper, and will be presented in a poster in a departmental mini-conference. Suggested project ideas will be provided by the instructors, but you are strongly encouraged to come up with your own project ideas. Potential projects include implementation or analysis of some piece of an OS, a DBMS, or an internet service; extending one of these systems with new functionality; or measurement and analysis of existing systems with the goal of better understanding issues in system design.

Overall Grading:

50% project paper
15% project demo
25% midterm
10% paper summaries

Lecture and Reading Schedule

There is no required textbook, but some of the papers are in Readings in Database Systems. We will read and discuss 2-4 papers per week. Most of the papers for the class will be available either on-line, or as handouts in a previous class. An additional suggested text is Gray and Reuter's Transaction Processing: Concepts and Techniques

Paper Summaries

For every assigned paper (except for some explicit exceptions), students must turn in a reading summary before the beginning of the corresponding lecture. Summaries should be e-mailed to cs262profs@db.cs.berkeley.edu. Summaries should be limited to 1/2 page at most and should include at least one criticism of the paper. (Just a simple summary, as you might get from the abstract or conclusion, misses the point, which is to get students to read the papers!) You can miss 3 summaries without penalty.

Part 1: Basics and Classics
Wed 8/27 Week 1	The UNIX Time-Sharing System [no summary required] Dennis M. Richie and Ken Thompson (New electronic version)
Mon 9/1 Week 2	Labor Day Holiday
Wed 9/3	A History and Evaluation of System R [in the Red Book] Donald D. Chamberlin, Morton A. Astrahan, Michael W. Blasgen, James N. Gray, W. Frank King, Bruce G. Lindsay, Raymond Lorie, James W. Mehl, Thomas G. Price, Franco Putzolu, Patricia Griffiths Selinger, Mario Schkolnick, Donald R. Slutz, Irving L. Traiger, Bradford W. Wade and Robert A. Yost Optional Reading 1: Architecture of a Database System [also in textbook] Optional Reading 2: The Design and Implementation of Ingres [in textbook] Michael Stonebraker, Eugene Wong, Peter Kreps and Gerald Held.
Part 2: Persistent Storage
M 9/8 Week 3	A Fast File System for UNIX McKusick, Joy, Leffler and Fabry Analysis and Evolution of Journaling File Systems Optional reading: The Design and Implementation of a Log-Structured File System Rosenblum and Ousterhout (229K)
W 9/10	The HP AutoRAID Hierarchical Storage System [My temporary local copy,\ 2-up version] Wilkes, Golding, Staelin and Sullivan
M 9/15 Week 4	ARIES: A Transaction Recovery Method Supporting Fine-Granularity Locking and Partial Rollbacks Using Write-ahead Logging , 2-up version C. Mohan et al. (in Red Book)
W 9/17	Stasis: Flexible Transactional Storage Russell Sears and Eric Brewer Guest lecture by Rusty Sears
M 9/22 Week 5	Lightweight Recoverable Virtual Memory M. Satyanarayanan, Henry H. Mashburn, Puneet Kumar, David C. Steere, and James J. Kistler
Part 3: Concurrency
W 9/24	Experience with Processes and Monitors in Mesa Butler Lampson and David Redell
M 9/29 Week 6	SEDA: AnArchitecture for Well-Conditioned, Scalable Internet Services Matt Welsh, David Culler, and Eric Brewer Capriccio: Scalable Threads for Internet Services Rob von Behren, Jeremy Condit, Feng Zhou, George C. Necula, and Eric Brewer Optional (no summary): Comparing the Performance of Web Server Architectures Pariag, Brecht, Harji, Buhr and Shukla
W 10/1	Granularity of Locks and Degrees of Consistency in a Shared Database Gray et al. (Also in the Red Book) On Optimistic Methods for Concurrency Control Kung and Robinson Optional reading: Generalized Isolation Levels Atul Adya, Barbara Liskov, Patrick O'Neil
M 10/6 Week 7	Concurrency Control Performance Modeling: Alternatives and Implications Agrawal et al.
W 10/8	Lottery Scheduling: Flexible Proportional-Share Resource Management Waldspurger and Weihl Optional reading: Stride Scheduling: Deterministic Proportional-Share Resource Management
Part 4: Query Processing
M 10/13 Week 8	Access Path Selection in a Relational Database Management System [all new version!] [in Red Book] Selinger, Astrahan, Chamberlain, Lorie & Price Optional reading: Grammar-like Functional Rules for Representing Query Optimization Alternatives G. Lohman [in Red Book]
W 10/15	Parallel Database Systems: The Future of High Performance Database Systems Dave DeWitt and Jim Gray MapReduce: Simplified Data Processing on Large Clusters Dean and Ghemawhat.
M 10/20 Week 9	Encapsulation of Parallelism Graefe Eddies: Continuously Adaptive Query Processing Ron Avnur and Joe Hellerstein
Part 5: Virtual Machines
W 10/22	Xen and the Art of Virtualization P. Barham, B. Dragovic, K Fraser, S. Hand, T. Harris, A. Ho, R. Neugebauer, I. Pratt and A. Warfield. Are Virtual Machine Monitors Microkernels Done Right? S. Hand, A. Warfield, K. Fraser, E. Kotsovinos, D. Magenheimer.
M 10/27 Week 10	Live Migration of Virtual Machines C. Clark, K. Fraser, S. Hand, J. Hansen, E. Jul, C. Limpach, I. Pratt, A. Warfield ReVirt: Enabling Intrusion Analysis through Virtual Machine Logging and Replay G. Dunlap, S. King, S. Cinar, M. Basrai and P. Chen
Part 6: Networking
W 10/29	Congestion Avoidance and Control Van Jacobson TCP Congestion Control with a Misbehaving Receiver Savage, Cardwell, Wetherall and T. Anderson
M 11/3 Week 11	Active Messages: A Mechanism for Integrated Communication and Control [no summary required] von Eicken, Culler, Goldstein, and Schauser U-Net: A User-Level Network Interface for Parallel and Distributed Computing Basu, Buch, Vogels, and von Eicken
W 11/5	The Click Modular Router Kohler, Morris, Chen, Jannotti and Kaashoek
Part 7: Distributed Computing
M 11/10 Week 12	Chord: A Scalable Peer-to-peer Lookup Service for Internet Applications Stoia, Morris, Karger, Kaashoek, Balakrishnan The Impact of DHT Routing Geometry on Resilience and Proximity Gummadi, Gummadi, Gribble, Ratnasamy, Shenker, Stoica
W 11/12	Lessons from Giant-Scale Services Brewer
M 11/17 Week 13	K42: Building a Complete Operating System Krieger et al.
W 11/19	Implementing Declarative Overlays Loo, Condie, Hellerstein, Maniatis, Roscoe, and Stoica
M 11/24 Week 14	Paxos Made Simple Lamport Late addition: two-phase commit [this is required but we will accept late summaries] Optional Wikipedia entry on Paxos
Part 8: Potpourri
W 11/26	Singularity: Rethinking the Software Stack G. Hunt and J. Larus Deconstructing Process Isolation [Singularity] M. Aiken, M. Fahndrich, C. Hawblitzel, G. Hunt; J. R. Larus
M 12/1 Week 15	Speculative Execution in a Distributed File System E. B. Nightengale, P. M. Chen and J. Flinn
W 12/3	Dynamo: Amazon's Highly Available Key-Value Store
M 12/8 Week 16	Beyond Pilots: Keeping Rural Wireless Networks Alive S. Surana, R. Patra, S. Nedevschi, M. Ramos, L. Subramanian, Y. Ben-David and E. Brewer No summaries required
W 12/10	Final Lecture

Lecture Notes

12/1: Speculator
11/26: Singularity
11/24: Paxos, notes from Joe Hellerstein
11/17: K42
11/12: Giant-scale services, CAP theorem proof, Amazon CTO Werner Vogels talking about CAP Theorem and Amazon's insfrastructure (minute 36+)
11/10: DHTs
11/5: Click Modular Router
11/3: U-net and Active Messages
10/29: TCP (the OS perspective)
10/27: Novel uses of VMs
10/23: Here is one paper with the "tire tracks" figure that show the time lag between innovation and $1B markets; it is a worthwhile read also about CS funding patterns.
10/22: Virtual Machines, Part 1 , Disco journal paper, paper on architectural requirements for enabling virtual machines (from 1974)
10/20: Exchange and Eddies
10/15: Parallel databases & Map-Reduce
10/13: Query optimization
10/8: Scheduling
10/6: Concurrency performance comparison
10/1: Concurrency Control
9/29: Threads & Events
9/24: Mesa , Needham Lauer paper on the duality between events (message handlers) and threads (with shared memory and locks)
9/22: LRVM
9/17: Rusty Sear's guest lecture notes
9/15: Aries slides
9/10: AutoRAID, Hitachi vertical bit animation
9/8: Filesystems
9/3: System R

8/27: Unix
The "second system" syndrome that we discussed:

An architect's first work is apt to be spare and clean. He knows he doesn't know what he's doing, so he does it carefully and with great restraint As he designs the first work, frill after frill and embellishment after embellishment occur to him. These get stored away to be used "next time." Sooner or later the first system is finished, and the architect, with firm confidence and a demonstrated mastery of that class of systems, is ready to build a second system. This second is the most dangerous system a man ever designs. When he does his third and later ones, his prior experiences will confirm each other as to the general characteristics of such systems, and their differences will identify those parts of his experience that are particular and not generalizable. The general tendency is to over-design the second system, using all the ideas and frills that were cautiously sidetracked on the first one. The result, as Ovid says, is a "big pile." -- Frederick P. Brooks, Jr. The Mythical Man-Month

CS262a:Advanced Topics in Computer Systems

Mon/Wed 1-2:30 in 306 SodaCCN: 26800