CS 607

Hot Topics in Systems

Fall 2024

Course Information

Have you ever wondered what the hot topics in systems are? What research problems interest top system conferences (SOSP, OSDI, NSDI, EuroSys, ATC, VLDB, SIGMOD)? What are the types of papers that are getting published at these conferences?
In this world of AI/ML, every major company (The Big Seven) is still looking for system designers and researchers. What systems do these companies work on? How do they design these systems? Do you want that edge during the interviews?
In this course, we will explore some of the latest research areas in systems, reviewing the most cited works and gathering insights behind those systems.

Time: Tues 2pm - 3:20pm PST
Location: Deschutes 200

Instructor

Name	Office Hours	Location
Suyash Gupta	By Appointment	Deschutes 334

Syllabus

Objectives

This is an upper-level course where we will study and discuss research papers from top systems, database, networking, and security conferences. This course places a heavy emphasis on attendance and paper reading/discussions. Additionally, each student will work towards coming up with a new research idea, which can lead to a future publication. Upon successful completion of this course, a student will have broader understanding in the following topics:

• Replicated State Machines.
• Shared-Disaggregated Memory.
• Vector Databases.
• Fault Tolerance.
• Federated Learning.

Grading Scheme

The final grade for the course will be based approximately on the following weights:

• 60% - Paper Discussions and Summary
• 40% - Research Idea

Papers

Shared-Disaggregated Memory

• Semeru: A Memory-Disaggregated Managed Runtime. OSDI 2020
• One-sided RDMA-Conscious Extendible Hashing for Disaggregated Memory. ATC 2021.
• Sherman: A Write-Optimized Distributed B+Tree Index on Disaggregated Memory. SIGMOD 2022
• FUSEE: A Fully Memory-Disaggregated Key-Value Store. FAST 2023.
• DiLOS: Do Not Trade Compatibility for Performance in Memory Disaggregation. EuroSys 2023.
• Memory disaggregation: why now and what are the challenges. ACM SIGOPS Oper. Syst. Rev. 2023.
• Motor: Enabling Multi-Versioning for Distributed Transactions on Disaggregated Memory. OSDI 2024.
• Understanding the Performance Implications of the Design Principles in Storage-Disaggregated Databases. SIGMOD 2024.
• LazyLog: A New Shared Log Abstraction for Low-Latency Applications. SOSP 2024.
• SWARM: Replicating Shared Disaggregated-Memory Data in No Time. SOSP 2024.

Vector Databases

• Survey of Vector Database Management Systems. VLDB Journal 2024
• Vector quantization. IEEE ASSP Magazine 1984.
• Efficient k-nearest neighbor graph construction for generic similarity measures. WWW 2011.
• Annoy: Approximate Nearest Neighbors Oh Yeah. Spotify.
• Lossless product quantization code compression for high dimensional similarity search. PVLDB 2020.
• A hybrid analytical engine towards query fusion for structured and unstructured data. PVLDB 2020.
• SPANN: Highlyefficient billion-scale approximate nearest neighbor search. NeurIPS 2021.
• Milvus: A purpose-built vector data management system. SIGMOD 2021.
• Manu: A cloud native vector database management system. PVLDB 2022.
• The design and implementation of a real time visual search system on JD e-commerce platform. Middleware 2018

Federated Learning

• Towards Federated Learning at Scale: System Design. SysML 2019.
• A survey on federated learning. Knowledge Based Systems 2021.
• Oort: Efficient Federated Learning via Guided Participant Selection. OSDI 2021.
• Walle: An End-to-End, General-Purpose, and Large-Scale Production System for Device-Cloud Collaborative Machine Learning. OSDI 2022.
• Papaya: Practical, Private, and Scalable Federated Learning. MLSys 2022.
• FedScale: Benchmarking Model and System Performance of Federated Learning at Scale. ICML 2022.
• Improving Federated Learning Personalization via Model Agnostic Meta Learning. Arxiv 2019.
• BatchCrypt: Efficient Homomorphic Encryption for Cross-Silo Federated Learning, ATC 2020.

Schedule

Date	Paper	Presenters
Oct 1	Introduction; Cornus: Atomic Commit for a Cloud DBMS with Storage Disaggregation.	Suyash Gupta
Oct 8	Spotify's Annoy	River, Quang, and Jonathan
Oct 8	Semeru: A Memory-Disaggregated Managed Runtime.	Zac and Nihal
Oct 15	Towards Federated Learning at Scale: System Design	Zac and Nihal
Oct 15	Oort: Efficient Federated Learning via Guided Participant Selection	River, Quang, and Jonathan
Oct 22	Sherman: A Write-Optimized Distributed B+Tree Index on Disaggregated Memory	River, Quang, and Jonathan
Oct 22	Walle: An End-to-End, General-Purpose, and Large-Scale Production System for Device-Cloud Collaborative Machine Learning	Zac and Nihal
Oct 29	Vector quantization	River, Quang, and Jonathan
Oct 29	Papaya: Practical, Private, and Scalable Federated Learning	Zac and Nihal
Nov 5	MidTerm Research Idea Discussions	TBD
Nov 12	TBD	TBD
Nov 19	TBD	TBD
Nov 26	TBD	TBD
Dec 3	Final Research Idea Discussions	TBD