Engineering Cloud Computing
ENGR-E 516/CSCI-B 649 (Fall 2023)
Announcements
Course Description
This course will teach the fundamental concepts, engineering principles, and practical skills pertaining to the effective use of cloud computing. This course will focus on both cloud applications and the design of cloud platforms. We will cover the relevant concepts from operating systems, computer networks, and distributed systems.
This course should be useful to anyone who wants a deeper understanding of how the cloud works, as well those who want to learn how to easily and effectively use the cloud for running their applications at low cost. We will look at a wide spectrum of cloud-based applications such as a parallel data processing (e.g., MapReduce), data storage and caching (e.g., key-value stores).
We will also look at the challenges involved in the efficient operation of large-scale cloud platforms with hundreds of thousands of servers. The course will cover a wide gamut of data center optimization techniques such as hardware virtualization, distributed resource management, and software-defined datacenters.
This course will expose students to popular cloud platforms such as Amazon EC2, Google Cloud Platform, Microsoft Azure, etc., and introduce students to new developments such as serverless computing and edge-clouds.
Prerequisites
The course has no official prerequisites. However, it requires a high comfort-level with systems programming and debugging. The assignments in this course will include nontrivial programming in the language of your choice. A good way to gauge your preparedness is to see how comfortable you are with the first programming assignment: Simple Key-Value Store. Spawn processes and sockets .
References
Text-books
- DS. Distributed Systems: Principles and Paradigms, 3rd Edition (Maarten Van Steen and Andrew Tanenbaum) Online version
Papers
Other references
- CCTP. Cloud Computing Theory and Practice. Dan C. Marinescu. (2nd edition)
- OS3EP : Operating Systems in Three Easy Pieces http://pages.cs.wisc.edu/~remzi/OSTEP/
Schedule, notes, and readings
Lecture | Topic | Slides | Reading |
---|---|---|---|
A0 | Course Intro | cloud/0-admin.pdf | Berkeley View |
A1 | Intro to cloud computing | cloud/1-intro-annot.pdf | |
A2 | ..continued | (same as above) | |
A3 | OS: system calls | cloud/2-OS-1.pdf | OS3EP Chapter 4 |
A4 | OS: concurrency | cloud/2-OS-annot.pdf | OS3EP |
A5 | Networks | cloud/3-net-2-annot.pdf | |
A6 | Networks: Socket programming | cloud/3-net-2-annot.pdf | |
F1 | Client-server modeling | cloud/4-servers-annot-1.pdf | Markov Chains |
F2 | -More queueing theory- | cloud/4-servers-annot2.pdf | M/M/1 Queues |
D1 | Hardware Virtualization | cloud/11-virt-1.pdf | |
D2 | CPU Virt | cloud/11-virt-2.pdf | 4. VMWare, 5. KVM |
D3 | Paravirtualization | cloud/11-virt-3.pdf | 3. Xen |
D4 | Memory Virtualization | cloud/11-virt-4.pdf | |
D5 | Live Migration | cloud/11-virt-5.pdf | 6. Xen-migration |
C1 | Cloud infrastructure | cloud/10-iaas.pdf | |
C3 | Cloud Storage | cloud/15-storage-annot.pdf | 11 |
Midterm | |||
C4 | OS Virtualization | cloud/13-osvirt.pdf | |
C2 | Functions as a Service | cloud/16-serverless-annot.pdf | 9, 10 |
D6 | Cluster management | cloud/12-clustmgmt-annot.pdf | 7. ESX, 8. Remus |
F3 | Parallel scaling | cloud/6-scaling-annot.pdf | Amdahl's Law |
F4 | Elastic scaling | cloud/6-scaling-annot.pdf | |
B1 | Map-Reduce | [[cloud/7-MapRed-annot-1.pdf cloud/7-MapRed-annot-2pdf.pdf | 1. MapReduce |
B2 | Spark | cloud/spark-annot.pdf | |
E2, E3 | Transient Computing | cloud/transient.pdf | 12. SpotCheck |
26 | Edge Computing | ||
27 | ML on clouds | ||
28 | Energy | Carbon-first SoCC and MGHPCC | |
Lightweight Virt | Firecracker and gVisor | ||
Course Wrapup |
Evaluation Criteria
Cloud computing is a fast evolving field. In the same spirit, the course is going to be fluid in its structure and evaluation, and also depend on student interest and capabilities. This is not a conventional "paint by numbers" course with structured homework etc.
The rough breakdown is as follows, but is subject to change:
Component | Weight |
---|---|
Programming assignments (4) | 40% |
Homework and Readings | 10% |
Midterm Exam | 20% |
Final exam | 20% |
Lecture notes and class participation | 10% |
Assignments
Students will implement various classic distributed algorithms (such as Map-Reduce, distributed key-value stores) on public clouds, and learn to use various cloud services such as Functions as a Service, various storage services, and how to use cloud VMs to develop and deploy applications.
The design oriented assignments will involve a large degree of programming and debugging. In most cases, the programming assignments are language agnostic (you can pick any reasonable programming language). However, you should be comfortable in atleast one systems programming language (C, C++, Rust, Go) for the final assignment.
A key learning objective of this course is to design, architect, and implement a distributed system from scratch, and to design useful test-cases for evaluating the implementation. Therefore, no starter-code or templates will be provided, to give students the maximum flexibility and freedom to explore the unconstrained design space. Points will be awarded for correct and faithful designs, complete implementation, adequate testing, and reports and documentation.
Most programming assignments will take significantly longer than you anticipate. Start early. Please see the assignment descriptions below (from last year), to get a sense of how they will look like. In general, all programming assignments in this course only specify the "end goal", and you must figure out how to get there: what and how to implement, what libraries to use, etc. There will be no starter-code, no templates, no training wheels. You are on your own. Apart from assignment 3, there are no group assignments.
Likely assignments and schedule:
# | Task | Approx Due Date |
---|---|---|
1 | Simple Key-Value Store. Spawn processes and sockets | Lec 5 |
2 | Deploy Assign 1 on GCP VMs using APIs | Lec 10 |
3 | Map-Reduce with functions | Lec 20 |
4 | LXC resource controller | End |
Exams
The exams will test how well students have understood various virtualization techniques, cloud performance and cost tradeoffs, and how techniques learnt in class can be applied to emerging cloud offerings and applications.
Participation and Lecture notes
Since a majority of the class instruction is on the whiteboard, each lecture will have two scribes who will prepare notes with the major concepts and questions covered in class.
Late submission policy
Students can avail a total of four late-days and use them as they wish. Beyond that, late submissions will not be accepted.
There is a tight integration of assignments and lectures. Hence, late submissions are discouraged. It is strongly recommended to start early—completing the assignments always takes more time than you think.
Michelin Star Grading
The grading in this course will favor students who turn in exceptional programs, reviews, and exam answers. Towards this end, we will use a "Michelin Star" system where points are awarded for high quality course products. Going over and beyond the standard evaluation criteria will fetch multiple stars. Students are eligible for an A (or A+) grade only if they have atleast one "star" across the course. Thus, it is not enough to turn in work that is merely correct. Students with three stars are automatically eligible for A/A+ grades irrespective of their performance in the rest of the course.
Examples of high-quality work
- Programs that are well documented, have a clean design, and implement something non-trivial in a clever way.
- Proofs that are correct and concise.
- Insightful and thoughtful paper reviews
- Exam answers that are crisp, insightful, and show a deep or unique understanding of the subject matter.
- A great question or answer during class discussions/office hours
Frequently Asked Michelin Star Questions/Clarifications
- Can I get a Michelin Star for my homework?
- If you have to ask, probably not. Not all assignments are eligible. For example, the preparatory assignments (1,2) are pretty basic, and leave little scope for ample creativity and excellence.
- If you are struggling in the course, please do not rely on the hope that later michelin stars will magically rescue your grade.
- In some cases, assignments will list extra components that will make the submission eligible.
- Your first priority should be to submit correct and complete assignments. Please do not over-design or over-engineer your assignments, lest you are unable to submit on time. A non-functional submission with a "great" design will likely get 0. Thus, be careful when shooting for the stars.
Administrative Information
Class Information
Where | When |
---|---|
Luddy Hall Room 4063 | Mondays and Wednesdays 9:45–11:00 |
Office Hours
Who | Office Location | Office Hours | |
---|---|---|---|
Prateek Sharma | prateeks | Luddy 4126 | Wed 4–5 |
Prabhat Suman | prsuman <at> iu | Luddy 4017 | Mon 3–4 |