Schedule and Reading List

Date	Readings	Presenter
Aug 25	Introduction + Systems Research	Marco
Aug 28	Reading/Reviewing Papers	[video]
Sep 1	Research Patterns	[video]
Sep 4	How to Innovate + Clear/Creative Thinking	[video]
Sep 8	Giving a Talk	[video]
Sep 11	Tech Trends in the Datacenter	[video]
	DC Topics
Sep 15	Arrakis: The Operating System is the Control Plane (Companion) IX: A Protected Dataplane Operating System for High Throughput and Low Latency
Sep 18	FaRM: Fast Remote Memory (Companion) Empowering Azure Storage with RDMA	Omar
Sep 22	No class - holiday
Sep 25	Datacenter RPCs can be General and Fast (Companion) StRoM: Smart Remote Memory
Sep 29	Homa: A Receiver-Driven Low-Latency Transport Protocol Using Network Priorities (Companion) Swift: Delay is Simple and Effective for Congestion Control in the Datacenter
Oct 2	Shenango: Achieving High CPU Efficiency for Latency-sensitive Datacenter Workloads (Companion) Shinjuku: Preemptive Scheduling for µsecond-scale Tail Latency	Majed
Oct 6	How to diagnose nanosecond network latencies in rich end-host stacks (Companion) Collie: Finding Performance Anomalies in RDMA Subsystems	Ibraheem
Oct 9	Offloading distributed applications onto smartNICs using iPipe (Companion) RDMA is Turing complete, we just did not know it yet!	Achref
Oct 13	No class - mid-semester break
Oct 16	Time Management + People Skills	[video]
Oct 20	Visualization and Plotting	Marco
Oct 23	Ensō: A Streaming Interface for NIC-Application Communication (Companion) PRISM: Rethinking the RDMA Interface for Distributed Systems
Oct 27	Mid-Semester Presentations
	ML systems
Oct 30	Rethinking Machine Learning Collective Communication as a Multi-Commodity Flow Problem (Companion) TACCL: Guiding Collective Algorithm Synthesis using Communication Sketches
Nov 3	MCCS: A Service-based Approach to Collective Communication for Multi-Tenant Cloud (Companion) Towards Domain-Specific Network Transport for Distributed DNN Training
Nov 6	Enabling Tensor Language Model to Assist in Generating High-Performance Tensor Programs for Deep Learning (Companion) Ansor: Generating High-Performance Tensor Programs for Deep Learning
Nov 10	Unicron: Economizing Self-Healing LLM Training at Scale (Companion) MegaScale: Scaling Large Language Model Training to More Than 10,000 GPUs
Nov 13	DistServe: Disaggregating Prefill and Decoding for Goodput-optimized Large Language Model Serving (Companion) Efficient Memory Management for Large Language Model Serving with Paged Attention
Nov 17	Fairness in Serving Large Language Models (Companion) Companion: S-LoRA: Serving Thousands of Concurrent LoRA Adapters
Nov 20	SpotServe: Serving Generative Large Language Models on Preemptible Instances (Companion) Parcae: Proactive, Liveput-Optimized DNN Training on Preemptible Instances	Jihao
Nov 24	Enabling Parallelism Hot Switching for Efficient Training of Large Language Models (Companion) Tenplex: Dynamic Parallelism for Deep Learning using Parallelizable Tensor Collections
Nov 27	Final Presentations
Dec 1	No class
Dec 4	No class