CMSC 838P Spring 2003 Internet-Scale Software Engineering: Leveraging Fielded Resources Instructor Adam Porter CSI 3118 TuTh 9:30pm-10:45pm

Class requirements will primarily be a mid-term, a final, active class participation (including at least one presentation), and a research project. The results of the research project will be captured in a publishable-quality report. 

Click here for weekly Reading Assignments

Click here for instructions on writing and submitting reviews for each reading assignment

Click here for reading schedule.

Dynamic analyses, such as testing, performance monitoring, and usage profiling, play a key role in state-of-art approaches to software development and quality assurance. With rare exception, these analyses are performed in-house, on developer platforms, using developer-provided input workloads. A benefit of in-house analyses is that programs can be instrumented and analyzed at very fine levels of detail since the developers have unrestricted access to the software. The shortcomings are severe, however, including:

• increased cost and schedule for development and QA activities, especially when software is released in several configurations and for several platforms;
• misleading results when the input workload differs from actual workloads or when the in-house system or execution environment differs from that found in the field; and
• limitations on the computing platforms and the number of computing cycles that are available for the analyses.

The situation has only worsened with the growth of the Internet and the emergence of ubiquitous computing. First, the widespread use of computer systems has caused a dramatic increase in the demand for software, forcing many companies to shorten their software development time and to release software without performing adequate analysis and testing. Second, many of today's software products are run in very complicated and varying environments, and it is often impractical to analyze and test these software products under all possible runtime environments and configurations before releasing them.

There is a need to develop techniques and tools to bridge the existing gap between the way software is analyzed in-house and the way it is used in the field. We believe that those techniques and tools will save a great deal of time, effort, and resources throughout the software industry.

Consequently, we will explore how to achieve the following objectives:

• Devise technologies and tools that help improve software cost, quality, and performance incrementally.
• Leverage resources in the field---in terms of both users and computing platforms---to gather data needed by these technologies and tools, thereby shortening the gap between in-house and in-the-field performance.
• Minimize human effort by automating analysis steps possible and by using state-of-the-art statistical to minimize analysis efforts.
• Minimize unnecessary end-user overhead, without their privacy and security

This course will be logically divided into two, roughly-equal, sections.

• The first section will be a basic introduction to the field of software engineering.  Consequently, I will not assume (nor require) that students have a background in software engineering before taking this class.  We will discuss the tasks that make up the software lifecycle. These include: requirements analysis, design, coding, testing, and evolution. Finally, we will survey current research in each of these areas.
• The second section will examine recent research aimed at improving some software lifecycle activity. In particular we will focus on how these activities must change to support internet-scale software development.

This course will count towards the various comprehensive exam requirements. There will be a midterm, a final, and a term paper/project.-