Course Description

Data management systems and technologies have historically played a pivotal role in the context of computing environments that involve large volumes of data and information. In fact, data base management systems (DBMS) are the critical components of most data-intensive application infrastructure. Furthermore, the underlying technologies, both in terms of language and query models as well as with respect to the system architectures, have reached a level of maturity that has enabled its use as a plug-and-play component without the need for detailed learning of its internals.

Recently, however, the entire area of data-management especially as it pertains to large-scale data arising from Internet and Web-based applications is at the cross-roads. The main question in this debate is the effectiveness of old DBMS paradigms: declarative query languages, independence of logical and physical data model, and the computational framework based on the Transaction concept. Several of the large Internet companies such as Google, Yahoo, and Amazon have put forth competing solutions for both building data-intensive scalable applications over the Internet/Web as well as for large-scale data anlaytics.

During this quarter, we will begin a joint exploration to gain a deeper understanding to participate in this debate. In particular, the following topics will be covered:

• Data Management Solutions for Enterprise Applications: Formal Model of Correctness, the Transaction Model, Database Concurrency Control Protocols, Transaction Failures and Recovery, and Database Recovery Protocols.
• Data Management Solution for Internet Applications: Google's Application Stack: Chubby Lock Service, BigTable Data Store, and Google File System; Yahoo's key-value store: PNUTS; Amazon's key-value store: Dynamo; Correctness Semantics of key-value store and its impact on application development.
• Data Analysis Platforms for Enterprise Data Analytics: Online Analytical Processing, Data Warehouse Architectures, and the Data Cube Model.
• Large-scale Data Analytics in the Internet Context: The MapReduce Paradigm for large-scale data analysis, run-time system for supporting scalable and fault-tolerant execution of MapReduce programs, other paradigms: PigLatin and Hive, and parallel databases versus MapReduce

The detailed lecture organization for the course appears below.

Pre-requisites: CMPSC 170.

Required Textbook: Transactional Information Systems by Gerhard WEIKUM and Gottfried VOSSEN

Instructor: Divy Agrawal, agrawal AT cs.ucsb.edu

Office hours: TuTh 1:00PM - 2:00PM, 3117 Harold Frank Hall, and by appointment.

Teaching Assistant:

• Ceren Budak. Office hours:   TBD.

• Homework Assignments: 75 points
• Programming Project: 150 points
• Midterm Exam @ 75 points each
• No Final Exam
• Late assignments will not be accepted without proof of an emergency.
• Any act of dishonesty will result in an "F" for the course and a report to the associate dean of students. This includes any sharing/borrowing/copying/leveraging of computer code from any source unless explicitly specified.