CS 60038: Advances in Operating Systems Design Lecture 1: Course Introduction

Abhijit Mondal Sumitro Bhaumik Bishakh Ghosh Teaching Assistants Abhijit Mondal abhimondal@iitkgp.ac.in Office: Systems and Mobile Research Lab Sumitro Bhaumik sumitros@gmail.com Office: Systems and Mobile Research Lab Bishakh Ghosh ghoshbishakh@gmail.com Office: Research Scholar Lab (Annex)

Pre-requisite for this course Undergraduate Operating System Concepts Process and Threads Process life cycle Process Scheduling Process Synchronization (Semaphores) Memory Management (Memory mapping) Virtual Memory Management (Page tables, Demand Paging) File Systems, IO Systems An Assignment 0 has been posted in the course website to recap the core of the above concepts. Please solve them individually and independently.

What We’ll Cover In This Course Topic 1: OS Design  The UNIX Timesharing System (https://people.eecs.berkeley.edu/~brewer/cs262/unix.pdf)  Exokernel (https://pdos.csail.mit.edu/6.828/2008/readings/engler95exokernel.pdf)  Linux Kernel Architecture Process Management and Scheduling Memory management and virtual process memory  Interrupts Threading, Hyperthreading  Some recent research works on OS Design  Drowsy power management, SOSP 2015 Light-Weight Contexts: An OS Abstraction for Safety and Performance, OSDI 2016 HawkEye: Efficient Fine-grained OS Support for Huge Pages, ASPLOS 2019

What We’ll Cover In This Course Topic 2: Virtualization  Xen and Art of Virtualization (https://www.cl.cam.ac.uk/research/srg/netos/papers/2003-xensosp.pdf)  Memory Management in Virtualization (Example: VMWare) - https://www.vmware.com/pdf/usenix_resource_mgmt.pdf Lightweight virtualization - http://www.cs.unc.edu/~porter/pubs/p291-porter.pdf Container Virtualization  Some recent research works on virtualization  My VM is Lighter (and Safer) than your Container, SOSP 2017 NEVE: Nested Virtualization Extensions for ARM, SOSP 2017

What We’ll Cover In This Course Topic 3: File System  Unix File System - https://people.eecs.berkeley.edu/~brewer/cs262/FFS.pdf Log Structured File System - https://people.eecs.berkeley.edu/~brewer/cs262/LFS.pdf Networked File System (NFS) - http://www.cs.ucf.edu/~eurip/papers/sandbergnfs.pdf Data Center File System (Google File System) - https://static.googleusercontent.com/media/research.google.com/en//archive/gfs- sosp2003.pdf Some recent researches on file systems Strata: A Cross Media File System, SOSP 2017 Pocket: Elastic Ephemeral Storage for Serverless Analytics, OSDI 2018

What We’ll Cover In This Course Topic 4: Some More Advanced and Recent Topics Trusted Execution Environments, ARM TrustZone System aspects of Blockchain -- Distributed Consensus  Automated Log Analysis  Machine learning on System Research  More Research Papers SeClaok: ARM Trustzone-based Mobile Peripheral Control, MobiSys 2018 Canopy: An End-to-End Performance Tracing And Analysis System, SOSP 2017 XFT: Practical Fault Tolerance beyond Crashes, OSDI 2016

Recent research papers from top OS conferences and journals Reading Materials "Professional Linux Kernel Architecture", Wolfgang Mauerer -- https://cse.yeditepe.edu.tr/~kserdaroglu/spring2014/cse331/termproject/B OOKS/ProfessionalLinuxKernelArchitecture-WolfgangMauerer.pdf Recent research papers from top OS conferences and journals ACM Symposiums on Operating Systems Principles (SOSP) Usenix Operating Systems Design and Implementation (OSDI) Architecture Support for Programming Languages and Operating Systems (ASPLOS) ACM Transactions on Computer Systems ACM Transactions on Storage Relevant materials will be published in the course website periodically

What You Can Expect from the Course In-depth knowledge of the very recent topics in the design of modern operating systems Virtualization technologies Distributed storage architecture (Ex. Google File System) Blockchain based system developments Embedded system concepts (ARM TrustZone) Hands-on experience of various system platforms Kernel compiling and debugging Handling hypervisors for VM setup and optimization Container management - Kubernetes Blockchain management -- Hyperledger platform

What the Instructor Expects from You Attending the classes regularly and making the classes interactive Feel free to stop me whenever you have a question in your mind You’ll recap the basic concepts of undergraduate OS course You’ll solve the assignments on time and by your own -- take help of the TAs (also the instructors) if required. Keep on browsing the recent research papers published in conferences like SOSP, OSDI, ASPLOS etc. (Check DBLP) Regularly participate in the term projects (we’ll discuss later in details). Make your hand dirty by exploring the “nasty” codes of Linux Kernel

Evaluation and Grading Mid Semester: 25% End Semester: 35% Term Project (TP): 25% Assignments: 15% Mid and end semester examinations will contain conceptual questions. You may score very high just by Participating in the classes Solving the assignments Expected course load per week -- 3 hours (Classes) + 1 hour (Assignments) + 1 hour (Reading) + 3 hours (TP) + 1 hour (Meeting with the mentor for TP)

Every team will be assigned a topic and a mentor Term Projects We’ll form some teams -- number of team members will be decided based on the number of students registered in the course Every team will be assigned a topic and a mentor The topics are broadly some parts of the current system researches we are doing at our lab Remote memory management Container state management Exploring Trusted Execution Environments (TEE) OS state management at distributed environments Blockchain based system management

Term Projects The objective will be to get a hands-on experience of recent system researches Meet weekly with your mentor to discuss about the project The team will divide the tasks among the team members, every member of the team is expected to contribute in the project -- this is crucial for evaluation If you work regularly and mark significant contribution at the project, you can expect a research paper out of it at some good system conference.

To Start With ... How to Read a Paper - http://ccr.sigcomm.org/online/files/p83-keshavA.pdf How (and How Not) to Write a Good Systems Paper - https://www.usenix.org/legacy/publications/library/proceedings/dsl97/good _paper.html

Use a three-pass approach First Pass (Quick Scan): How to Read a Paper Use a three-pass approach First Pass (Quick Scan): Read title, abstract, introduction Read the section and sub-section headings ignoring details Read conclusion Check the references Get a view of the category, context, contribution and clarity of the paper

Take down notes whenever necessary How to Read a Paper Second Pass: Check the figures, diagrams etc. Check the unread references from the paper -- good way to know the background (only give the first pass) Third Pass: Give attention to the detailed content of the paper Take down notes whenever necessary Try to judge the paper critically -- you can even get new nice ideas (you can pass these ideas to me, will be happy to discuss and follow up) Why do you think that the paper got accepted? Try to write a review of the paper (some of these will be your assignment  )

Some Fundamentals - Process Management Consider a single-processor system. The process scheduler schedules the processes in a multi-tasking environment. How is this process scheduler implemented? Is it a software or a hardware? Who schedules the process scheduler?

Some Fundamentals - Memory Management What is a virtual memory? What is the basic philosophy behind the design of virtual memory principles?