CIS 700-5: The Design and Implementation of Cloud Networks

Spring 2017

NOTE: Class will be held in the DSL Conference Room, NOT the room listed in Penn InTouch

Instructor: Vincent Liu
Room: Distributed Systems Laboratory (DSL) Conference Room
Time: Monday and Wednesday 1:30-3:00 PM
Discussions: Piazza
Office hours: Wednesday 3-4 PM (470 Levine North) or by appointment

Course Description

Much of today's computing occurs in and around data centers. Large, specialized networks connect each data center's servers together and also connect those servers to the outside world. This graduate seminar course will explore the design and implementation of these massive network deployments---inside data centers, between data centers, and in the Internet.

A crucial part of the exploration is understanding how large cloud networks are built today by companies like Facebook, Google, and Microsoft. Equally important is where recent research is taking them. This seminar will cover topics such as network architecture, congestion control, troubleshooting, OS/application design, and software-defined networking.

This course satisfies a PhD seminar requirement for PhD students.

Prerequisites

This class is geared toward PhD students or Masters students that are interested in ongoing research. General background in computer systems and networking is preferred.

Format

The class will meet for 1.5 hours twice per week. The format is largely discussion based. Every week, students will be expected to read a couple papers and come to class prepared to discuss them. All students will need to submit an interesting thought about the readings before class. Additionally, one student will be tasked with leading discussion for the day.

In addition to the reading and discussions, students will, in groups of 2-3, engage in a research project of their choosing. The hope is that the project will bridge cloud networking and students' existing research/interests. More concretely, students will be expected to write a proposal for their chosen research project, and produce a presentation and written report of their results at the end of the semester.

Grading

Reading and participation: 50%

Paper reviews: 20%
Leading discussion: 15%
Discussion participation: 15%

Project: 50%

Schedule

Date	Topic	Discussion Leader	Reading List
Jan. 11	Introduction [PPT]	Vincent	Optional: S. Keshav, How to read a paper
Jan. 16	No Class (Martin Luther King, Jr. Day)
Jan. 18	Data Center Architectures [PPT]	Vincent	A Scalable, Commodity Data Center Network Architecture VL2: A Scalable and Flexible Data Center Network
Jan. 23	Network Topologies [PPT]	Vincent	F10: A Fault-Tolerant Engineered Network Jellyfish: Networking Data Centers Randomly
Jan. 25	Data Center Traffic Patterns	Josh Fried	Network Traffic Characteristics of Data Centers in the Wild Inside the Social Network's (Datacenter) Network
Jan. 30	Optical/Wireless Networks	Sam Weinberg	Helios: A Hybrid Electrical/Optical Switch Architecture for Modular Data Centers Augmenting Data Center Networks with Multi-Gigabit Wireless Links
Feb. 1	Congestion Control	Ji Yong Cho	Data Center TCP (DCTCP) TIMELY: RTT-based Congestion Control for the Datacenter
Feb. 6	No Class (Google Summit)
Feb. 8	No Class (Google Summit) *Project proposals/plan due.*
Feb. 13	Traffic Balancing	Sam Weinberg	Improving Datacenter Performance and Robustness with Multipath TCP CONGA: Distributed Congestion-Aware Load Balancing for Datacenters
Feb. 15	Load Balancing	Qizhen Zhang	Maglev: A Fast and Reliable Software Network Load Balancer Ananta: Cloud Scale Load Balancing
Feb. 20	VMs and Containers	Mihir Pattani	Network Virtualization in Multi-tenant Datacenters XLDB2015: Kubernetes and the Path to Cloud Native
Feb. 22	Kernel Bypass	Henri Maxime Demoulin	Arrakis: The Operating System is the Control Plane RDMA over Commodity Ethernet at Scale
Feb. 27	Multitenant Networking	Henri Maxime Demoulin	Application-Driven Bandwidth Guarantees in Datacenters
Mar. 1	Scheduling	Saeed Abedi	Mesos: A Platform for Fine-Grained Resource Sharing in the Data Center Large-scale cluster management at Google with Borg
Mar. 6	No Class (Spring Break)
Mar. 8	No Class (Spring Break)
Mar. 13	Software-defined Networking (SDN)	Mihir Pattani	Talk: Gentle Introduction to Software-Defined Networking Frenetic: A Network Programming Language
Mar. 15	Wide-area SDN	Ji Yong Cho	B4: Experience with a Globally-Deployed Software Defined WAN Achieving High Utilization with Software-Driven WAN
Mar. 20	Programmable switches	Mikael Mantis	P4: Programming Protocol-Independent Packet Processors Packet Transactions: High-Level Programming for Line-Rate Switches
Mar. 22	Network Function Virtualization (NFV)	Akshay Sriraman	ONS2014 Keynote: Amin Vahdat OpenNF: Enabling Innovation in Network Function Control
Mar. 27	No Class (NSDI)
Mar. 29	No Class (NSDI)
Apr. 3	Cloud Applications	Qizhen Zhang	CODA: Toward Automatically Identifying and Scheduling Coflows in the Dark CherryPick: Adaptively Unearthing the Best Cloud Configurations for Big Data Analytics
Apr. 5	Cloud Security	Josh Fried	Shielding Applications from an Untrusted Cloud with Haven Building Web Applications on Top of Encrypted Data Using Mylar
Apr. 10	Ordering	Saeed Abedi	Globally Synchronized Time via Datacenter Networks Just Say NO to Paxos Overhead: Replacing Consensus with Network Ordering
Apr. 12	Monitoring and Diagnostics	Mikael Mantis	Packet-Level Telemetry in Large Datacenter Networks One Sketch to Rule Them All: Rethinking Network Flow Monitoring with UnivMon
Apr. 17	Network Management	Akshay Sriraman	Robotron: Top-down Network Management at Facebook Scale Evolve or Die: High-Availability Design Principles Drawn from Google's Network Infrastructure
Apr. 19	No Class (Finish Projects)
Apr. 24	Project Presentations
Apr. 26	Project Presentations