Course Description

This course explores advanced topics in highly scalable Internet services and their underlying systems architecture. Software today is increasingly being delivered as a service: accessible globally via web browsers and mobile applications and backed by millions of servers. Modern frameworks and platforms are making it easier to build and deploy these systems, such as Ruby on Rails and Amazon’s EC2.

Yet despite these advances, some concerns just don’t go away. Building scalable Internet services today still requires an understanding of topics like caching, load balancing, security, and monitoring. In this course we will examine these topics and more: the state of the art in building scalable Internet services.


Course lectures will be Tuesdays and Thursdays, from 1:00pm to 2:50pm in Phelps 2510. Lecture material will cover the essentials of building large scale Internet services. The lecture schedule and slides are available online.


The goal of the course project is to gain hands-on experience in building and deploying a scalable web service on the Internet. Students will do this using some of the latest web technologies in order to learn how to tackle scalability and fault-tolerance concerns. Projects will be conducted in agile teams of five students, and team will build their own scalable web site using fundamental web technologies and the Ruby on Rails framework.

Project-centric lab will meet every Tuesday from 4:00pm to 5:00pm in Phelps 3525. These mandatory labs offer a chance to collaborate with your teammates, demo your progress to the instructor, and get guidance.

Learning Outcomes

By the end of this course CS291A students will be able to:

  • create a Ruby on Rails web service and deploy it to Amazon’s EC2
  • detail a significant majority of the general services and protocols that are, or may be, involved when browsing to a high-traffic volume web service such as
  • determine via measurement where bottlenecks exist in a web service under their control
  • employ memcached to reduce the load on an uncached web service in order to handle a larger volume of traffic
  • utilize git to contribute to open source projects on github via pull requests


At the end of the quarter each group’s project will be assigned a grade based on their web service being of sufficient complexity, and the group’s description of their methodological approach to load testing and subsequent scaling via various techniques as described in their project write-up and/or in their project presentation. Objectively these components break down into:

  • 30%: web service complexity
  • 50%: load testing and scaling (communicated through presentation and/or write-up)
  • 10%: quality of project presentation
  • 10%: quality project write-up

On an individual basis, 5% of one’s grade is based on their participation in-class and on piazza.

The remaining 95% of an individual’s grade will be computed by multiplying their group’s project grade by an individual score. Individual scores are computed by suming the relative percent of work each individual confidentially assigns other individuals in their group.

For example, given a three person group with a group grade of 95%, everyone individually with 100% participation, and the following peer-graded scores:

  Alice Bob Chuck
Alice - 55% 45%
Bob 60% - 40%
Chuck 52% 48% -
Total 112% 103% 85%

Then Alice would end with a 106.4% (A+), Bob with 97.85% (A+) and Chuck with 80.75% (B-).

So that everyone knows where they stand within their group, we will confidentially conduct the peer grading process three times during the course (November 2, November 21, and December 7). Only the outcome of the final peer-grade (December 7) will be used to compute the final grade.

Letter Grades

Letter grades will be assigned as follows:

Percent Grade
96 ⅔ A+
93 ⅓ A
90 A-
86 ⅔ B+
83 ⅓ B
80 B-
76 ⅔ C+
73 ⅓ C
0 F

The number listed is the lower bound of the percent grade for the given range. It is expected that you are a relatively equal contributor in your group, with adjustments for significant positive outliers. Thus, please take note of the drop directly from C to F.