CIS 455 / 555: Internet and Web Systems (Spring 2013)

Instructor Andreas Haeberlen
Location: 560 Levine Hall North (a.k.a. GRW building)
Location Location: Moore 216
Mondays + Wednesdays 10:30am - noon
Teaching assistants Dhruv Arya, adhruv@seas.upenn.edu
Office hour: TBA

Santosh Kumar Balakrishnan, bsanth@seas.upenn.edu
Office hour: TBA

Ping Fu, fuping@seas.upenn.edu
Office hour: TBA

Saurabh Garg, saurabhg@seas.upenn.edu
Office hour: TBA

Chetan Singh, chetans@seas.upenn.edu
Office hour: TBA

Anand Subramanian, anandsub@seas.upenn.edu
Office hour: TBA

Course description This course focuses on the issues encountered in building Internet and web systems: scalability, interoperability (of data and code), atomicity and consistency models, replication, and location of resources, services, and data. Note that it is not about building database-backed or PHP/JSP/Servlet-based web sites (for this, see CIS 330/550 or MKSE 212). Here, we will learn how a Servlet server itself is built!

We will examine how XML standards enable information exchange; how web services support cross-platform interoperability (and what their limitations are); how "cloud computing" services work; how to do replication and Akamai-like content distribution; and how application servers provide transaction support in distributed environments. We will study techniques for locating machines, resources, and data (including directory systems, information retrieval indexing and ranking, web search, and publish/subscribe systems); we will discuss collaborative filtering and mining the Web for patterns; we will investigate how different architectures support scalability (and the issues they face). We will also examine the ideas that have been proposed for tomorrow's Web, including the "Semantic Web", and see some of the challenges, research directions, and potential pitfalls.

An important goal of the course is not simply to discuss issues and solutions, but to provide hands-on experience with a substantial implementation project. This semester's project will be a peer-to-peer implementation of a Googe-style search engine, including distributed, scalable crawling; indexing with ranking; and even PageRank. We will also incorporate the use of topic-specific recognizers and mash-ups.

As a side effect of the material of this course, you will learn about some aspects of large-scale software development: assimilating large APIs, thinking about modularity, reading other people's code, managing versions, debugging, and so on.

UPDATE: CIS555 is now a core course for the MSE degree. For details, please see the MSE requirements. Also, the Daily Pennsylvanian recently published a nice article about CIS455/555.

Format The format will be two 1.5-hour lectures per week, plus assigned readings from handouts. There will be regular homework assignments and a substantial implementation project with experimental validation and a report. There will also be a midterm and a final exam.
Prerequisites This course expects familiarity with threads and concurrency, as well as strong Java programming skills. Those highly proficient in another programming language, such as C++ or C#, should be able to translate their skills easily. The course will require a considerable amount of programming, as well as the ability to work with your classmates in teams.
Texts and readings Distributed Systems: Principles and Paradigms, 2nd ed, by Tanenbaum and van Steen, Prentice Hall
Additional materials will be provided as handouts or in the form of light technical papers.
Grading Homework 25%, midterm 15%, final exam 15%, project 40%, participation 5%.
Other resources We will be using Piazza for course-related discussions; please sign up here. A reading list is also available.
Assignments The homework assignments will be available here. If necessary, you can request an extension (requires PennKey login).
Final project Wondering what you will be able to do at the end of this class? Here is an example from Spring 2012:
   
Hitchhiker's homepage
Results for "Apple"
Dhruv, Santhosh, Saurabh, and Chetan
More search options
Page preview
Dhruv Arya, Shantosh Kumar Balakrishnan, Saurabh Garg, and Chetan Singh built "Hitchhiker" a cloud-based search engine. Hitchhiker consists of 1) a scalable distributed crawler that runs on Amazon EC2 instances and uses FreePastry for coordination; 2) an indexer and a PageRank engine that is based on Elastic MapReduce; and 3) a web frontend. Hitchhiker also contains a number of extra features, including page previews, a "safe search" to filter out explicit results, and a special search for the visually challenged, which enables the user to control the search entirely with spoken commands. Google donated four Nexus cell phones as a prize for the best project, and each member of the Hitchhiker team received one of the phones.
Schedule
Date Topic Details Reading Remarks
Jan 9 Introduction Principles of building systems
Project management & debugging tips
Lampson: Hints for Computer Systems Design  
Jan 14 Server architectures Common server types: Web, application
Architectures: client/server, P2P, multi-tier
Threads, monitors, signals, producer-consumer
Thread pools, event-driven programming
Marshall: HTTP Made Really Easy
Krohn: OKWS paper
 
Jan 16 Krishnamurthy/Rexford Chapter 4
Tanenbaum 3.1
Jan 21 MLK day -- no class
Jan 23 Naming & locating resources Naming and directories; search strategies
LDAP; DNS; DNSSEC
Wikipedia: DNS
Marshall: LDAP intro
 
Jan 28 Indexing Document indexing
B+ tree
Comer: The Ubiquitous B-Tree  
Jan 30 Representing data Data representations, schemas
JPEG, MP3, and QT
XML
XPath and XSLT
Doan, Halevy, Ives: XML  
Feb 4 XSLT Tutorial HW1 MS1 due
Feb 6 Decentralized systems Partly and fully decentralized systems
Key-based routing
Partitioning and consistent hashing
BitTorrent, Chord, Pastry
Druschel and Rodrigues: Peer-to-peer systems  
Feb 11 Stoica et al.: Chord  
Feb 13 Retrieving data Crawling basics
Publish-subscribe; collaborative filtering
Mercator; XFilter
Altinel and Franklin: XFilter
Heydon and Najork: High-Performance Web Crawling
 
Feb 18 Storing, distributing, retrieving, and processing data Cloud file system
MapReduce programming model
Ghemawat et al.: The Google File System
Dean and Ghemawat: MapReduce
Feb 20  
Feb 25 Storing, distributing, retrieving, and processing data Hadoop
HDFS
Shvachko: Apache Hadoop: The Scalability Update  
Feb 27 Midterm
Mar 4 Spring break -- no class
Mar 6
Mar 11 Code interoperability Remote procedure calls
Web services
SOAP, WSDL, REST
Service composition
XQuery
Tanenbaum chapters 4.2 and 10.3
SOAP tutorial
WSDL tutorial
Mar 13  
Mar 13 XQuery tutorial  
Mar 18 Documents and ranking Information retrieval models
Web connectivity
Ranking
Web crawlers
HITS and PageRank
Baeza-Yates Chapters 2 and 8
Kleinberg: HITS
Brin and Page: PageRank
Brin and Page: Google
Wired article on Google
Mar 20  
Mar 25 The Cloud Utility computing model
AWS basics; EC2+EBS
Armbrust: A view of Cloud Computing  
Mar 27 Transactions Application server and TP monitor architectures
ACID properties
Two-phase commit
Tanenbaum chapters 8.5-8.6  
Apr 1 Fault tolerance Replicated state machines
Consensus; Paxos algorithm
Rational behavior and Byzantine faults
Lamport: Paxos (Alternative version)
Schneider: State Machine Approach
 
Apr 3 Security Web security
Views, ACLs, capabilities; crypto basics
Kerberos; TLS
Tanenbaum chapter 9  
Apr 8 Incremental processing Bigtable
Percolator
Peng and Dabek: Percolator  
Apr 10 Special topics Accountability    
Apr 15 Differential privacy    
Apr 17 TBA    
Apr 22 Second midterm
  Project demos and reports      
Previous versions Spring 2012 |  Spring 2011 (taught by Zachary Ives until 2010)