Schedule and Reading List

Date	Readings	Presenter
Jan 27	Introduction	Marco
Jan 30
Feb 3	Background	Marco
Feb 6	Background	Marco
	Datacenter Networking and Transport
Feb 10	VL2: A Scalable and Flexible Data Center Network	Marco
Feb 13	Re-architecting datacenter networks and stacks for low latency and high performance (Companion) Data center TCP (DCTCP)	Marco
	Scheduling and Programming the Datacenter
Feb 17	Borg: Large-scale cluster management at Google with Borg also skim: Borg, Omega, and Kubernetes (Companion) Mesos: A Platform for Fine-Grained Resource Sharing in the Data Center
Feb 20	The Dataflow Model: A Practical Approach to Balancing Correctness, Latency, and Cost in Massive-Scale, Unbounded, Out-of-Order Data Processing (Companion) Ciel: a universal execution engine for distributed data-flow computing
Feb 24	The Scalable Commutativity Rule: Designing Scalable Software for Multicore Processors (Companion) IX: A Protected Dataplane Operating System for High Throughput and Low Latency
Feb 27	Size-aware Sharding For Improving Tail Latencies in In-memory Key-value Stores (Companion) Rein: Taming Tail Latency in Key-Value Stores via Multiget Scheduling
Mar 3	No class
	AI Systems
Mar 6	Democratizing Production-Scale Distributed Deep Learning Michelangelo FBLearner Flow
Mar 10	Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour (Companion) Large Scale Distributed Deep Networks
Mar 13	Hogwild!: A Lock-Free Approach to Parallelizing Stochastic Gradient Descent (Companion) Project Adam: Building an Efficient and Scalable Deep Learning Training System
Mar 17	Tiresias: A GPU Cluster Manager for Distributed Deep Learning (Companion) Optimus: an efficient dynamic resource scheduler for deep learning clusters
Mar 20	Ray: A Distributed Framework for Emerging AI Applications (Companion) MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems
Mar 24	Spring break
Mar 27	Spring break
Mar 31	No class (SysML)
Apr 3	No class (SysML)
Apr 7	Mid-Semester Presentations
Apr 10	Beyond Data and Model Parallelism for Deep Neural Networks	Marco
Apr 14	GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism (Companion) PipeDream: Fast and Efficient Pipeline Parallel DNN Training
Apr 17	Learning to Explain: An Information-Theoretic Perspective on Model Interpretation (Companion) The Mythos of Model Interpretability
Apr 21	Clipper: A Low-Latency Online Prediction Serving System (Companion) TensorFlow-Serving: Flexible, High-Performance ML Serving
Apr 24	Deep Learning with Limited Numerical Precision
Apr 28	In-Datacenter Performance Analysis of a Tensor Processing Unit (Companion) Direct Universal Access: Making Data Center Resources Available to FPGA
May 1	No class
	Correctness
May 5	Verdi: A Framework for Implementing and Formally Verifying Distributed Systems (Companion) IronFleet: Proving Practical Distributed Systems Correct
	Serveless Computing
May 8	SAND: Towards High-Performance Serverless Computing (Companion) Occupy the Cloud: Distributed Computing for the 99%
May 12	Wrap Up
May 15	Final Presentations