CPS 516: Data-intensive Computing Systems, Fall 2012

Announcements

• Project 2 is due on Nov 2 by 5.00 PM.
• Course lectures will be posted on the "Course schedule and notes" page. See link on the left.

Course Description

Database systems are going through very interesting and chaotic times. Popular relational database systems like IBM DB2, Microsoft SQLServer, Oracle, and Sybase are struggling to handle the massive scale of data introduced by the Web. Today, companies have to deal with extremely large datasets. Facebook absorbs 15 TeraBytes of data each day into their 2.5 PetaByte Hadoop-powered data warehouse. eBay maintains a 6.5 PetaByte (i.e., 6.5 x 1,000,000,000,000,000 Bytes!) data warehouse.

A new breed of database systems are emerging to handle data at massive scale. These systems take some of the successful features from conventional relational databases---like run-time query optimization, automated crash recovery, and self-tuning---and make them work at the scale of 100s-1000s of processors and disks. As we move into the world of "big data", many traditional assumptions break, new query and programming interfaces are required, and new computing models will emerge. Did you know that each Google search query can touch up to 2,000 servers that must all execute that query and respond in less than a third of a second?

This course covers a spectrum of topics from core techniques in relational data management to highly-scalable data processing using parallel database systems and MapReduce. The course material will be drawn from textbooks as well as recent research literature. The following topics will be covered this year. The figures in brackets indicate the amount of time devoted to the topic relative to the total duration of the course.

• Scalable data processing (40%)
  • MapReduce and systems based on MapReduce (Hadoop, Pig, Hive)
  • Parallel processing
  • Scalable key-value stores (Amazon Dynamo, Google BigTable, HBase)
  • Processing rapid, high-speed data streams
• Principles of query processing (35%)
  • Indexes
  • Query plans and operators
  • Cost-based query optimization
• Data storage (15%)
  • Databases Vs. FileSystems (Google FileSystem, Hadoop Distributed FileSystem)
  • Data layouts (row-stores, column-stores, partitioning, compression)
• Concurrency control and recovery (10%)
  • Consistency models for data (ACID, Serializability)
  • Write-ahead logging

Prerequisites: An introductory database course will be helpful, but it is not required. If you have not taken an introductory database course before, please talk to the instructor first. A lot of the material that we cover cannot be found in textbooks. Be prepared to do a fair amount of reading.

Time and Place

3:05pm-4:20 PM on Mondays and Wednesdays; in D106 LSRC.

Books and References

(Highly recommended) Hadoop: The Definitive Guide, by Tom White. O'Reilly Media. May 2012. (Third edition of the book at Amazon.com)

Database Systems: The Complete Book, by Hector Garcia-Molina, Jeffrey D. Ullman, and Jennifer Widom. Prentice Hall. 2002. (The second edition is available now.)

Readings will be posted on the readings page.

Staff

Instructor: Shivnath Babu
Office: D338 LSRC, Phone: 919-660-6579 (email is recommended)
Office hours: The instructor prefers to have office hours by appointment so that we make the best use of time. Send the instructor an email to fix the meeting time. The office hours will be held in the instructor's office.

TA: Jie Li
Office: D337 LSRC, Phone: 919-660-4011
Office hours and venue: Monday & Thursday 11:00-12:00 D337 LSRC

Grading

There is a semester-long course project (done in groups of two or three). Details will be presented in class.

Both midterm and final exams are open-book and open-notes. Laptops and other electronic devices are not allowed. Late homeworks will not be accepted, unless there are documented excuses from a physician or dean.

Honor Code

Under the Duke Honor Code, you are expected to submit your own work in this course, including homeworks, projects, and exams. On many occasions when working on homeworks and projects, it is useful to ask others (the instructor or other students) for hints or debugging help, or to talk generally about the written problems or programming strategies. Such activity is both acceptable and encouraged, but you must indicate in your submission any assistance you received. Any assistance received that is not given proper citation will be considered a violation of the Honor Code. In any event, you are responsible for understanding and being able to explain on your own all written and programming solutions that you submit. The course staff will pursue aggressively all suspected cases of Honor Code violations, and they will be handled through official University channels.