CSE291H is a graduate-level course about recent operating systems research, with a focus on datacenters. The course involves reading and discussing research papers and a research project. This quarter we will read papers about a variety of topics: multicore operating systems, network stacks, scheduling, memory management, disaggregation, and new devices such as SmartNICs, FPGAs, GPUs, and TPUs.
The goals of this course are:This course is open to PhD and Masters students as well as advanced undergraduate students. Students should have completed CSE 221 or an equivalent graduate-level operating systems course prior to enrolling.
Each class has 1-2 assigned papers. Students are expected to read the papers ahead of time, submit a short review about each paper (by 11:59 pm the evening before), and come to class prepared to discuss! Classes will be interactive and everyone is expected to participate.
Each student will lead the discussion of one paper. Students will share their discussion outline with the instructor at least two days before the discussion so that they can receive feedback on it.
The goal of this assignment is to familiarize students with CloudLab, so that they may use it as an experimentation platform for their research projects.
A major component of this course will be an open-ended research project, conducted individually or in groups of 2-3 students. Students will submit a brief project proposal and a final project write-up, and will also give an in-class presentation about their project at the end of the quarter.
| Date | Topics | Papers | Slides |
|---|---|---|---|
| Th 9/22 | Course overview | Intro | |
| Tu 9/27 | Multicore, intro to CloudLab | Multikernel (SOSP '09), CloudLab (ATC '19) - only first 2 sections | CloudLab, Multicore |
| Th 9/29 | Network stacks | IX (OSDI '14), XDP (CoNEXT '18) | Network stacks |
| Tu 10/4 | RDMA and RPCs | FaRM (NSDI '14) | RDMA and RPCs |
| Th 10/6 | RDMA and RPCs | eRPC (NSDI '19), PRISM (SOSP '21) | Datacenter networking |
| Tu 10/11 | Congestion control | Homa (SIGCOMM '18), Swift (SIGCOMM '20) | Datacenter congestion control |
| Th 10/13 | CPU scheduling | Killer Microseconds (CACM '17), Shenango (NSDI '19) | CPU scheduling |
| Tu 10/18 | CPU scheduling | ghOSt (SOSP '21) | CPU scheduling |
| Th 10/20 | Performance diagnosis | NSight (NSDI '22), Collie (NSDI '22) | none |
| Tu 10/25 | no class | ||
| Th 10/27 | Datacenter tax, SmartNICs | Warehouse-scale computers (ISCA '15), AccelNet (NSDI '18) | Processor pipelines |
| Tu 11/1 | SmartNICs | iPipe (SIGCOMM '19), nanoPU (OSDI '21) | SmartNICs |
| Th 11/3 | GPUs | PTask (SOSP '11) | GPUs |
| Tu 11/8 | TPUs | TensorFlow (OSDI '16) | TPUs |
| Th 11/10 | FPGAs | AmorphOS (OSDI '18), Coyote (OSDI '20) | FPGAs |
| Tu 11/15 | Disaggregation | LegoOS (OSDI '18) | Disaggregation |
| Th 11/17 | Memory management | Llama (ASPLOS '20) | Memory management |
| Tu 11/22 | Memory management | TLB shootdowns (EuroSys '20) | TLBs |
| Th 11/24 | Thanksgiving holiday | ||
| Tu 11/29 | Project presentations | ||
| Th 12/1 | Project presentations |