Grading rules

The final grade is computed according to the following rules:

• A+: A: [90%, 95%); A-: [85%, 90%)
• B+: [80%, 85%); B: [75%, 80%); B-: [70%, 75%)
• C+: [66%, 70%); C: [63%, 66%); C-: [60%, 63%)
• D+: [56%, 60%); D:[53%, 56%); D-: [50%, 53%)
• F: < 50%

 

Late Days

Each student will be granted 5 late days to use for the homework. After all free late days are used up, the penalty is 10% for each additional late day. Late days can not be used for the mid-way and final research presentation.

 

Student Integrity

Cheating in this course will result in a grade of F for the course and the University policies will be followed.

 

Students with Disabilities

If you need any accommodation, you are highly encouraged to contact both your instructor and Disability Resources Center (DRC).

 

Office hours

• Where: 4-205 Keller hall or Zoom
• When: Wednesday 10 -12 pm
• Additional office hours can be requested via Slack
  
  

Course Schedule

The course schedule is tentative and subject to change.

Lecture	Date	Topic	Required Readings	Optional Readings	Announcements
1	Tue: Sept 5	Introduction and Logistics	Armbrust2010, Vaquero11	Rackspace2012, Shafii2012, DeanSOSP2015, Cano2016, Vieira2012, Vogels2016, reiss2012, Ferguson2012, Rajagopalan2013, Das2013	Assignment 1: student info
2	Thu: Sept 7	Introduction to Cloud Computing	//	//
3	Tue: Sept 12	Virtualization I	VMware2007, Xen2005	VMware2006, VMware2007, Xen2003, Xen2005
4	Thu: Sept 14	Virtualization II	//	//	Proposal literature review due: Sept 26th Coding Assignment 1 due: Sept 30th
5	Tue: Sept 19	Containers I	Namespaces in operation, CGroups documentation	VM lighter than container, Slacker
6	Thu: Sept 21	Container II	Docker docs, Docker architecture	Improving Docker registry	Homework assignment 2: due date Oct 5th
	Tue: Sept 26	Project Ideas Discussion			Proposal and literature review due Today
7	Thu: Sept 28	Cloud Storage	GFS Paper, HDFS Paper	Hadoop Architecture Guide
8	Tue: Oct 3	Consistency Models	Visual Guide to NoSQL Systems, Delta Store (Meta), BespoKV paper	Chain Replication, CRAQ
9	Thu: Oct 5	Key-Value Stores	Scaling Memcached at Facebook, DynamoDB	//
10	Tue: Oct 10	Programming Model I	MapReduce, Spark	Hadoop vs Spark, MapReduce Architecture, Replex
11	Thu: Oct 12	Programming Model II	Pocket paper	A Berkeley View on Serverless Computing, A Berkeley View of Cloud Computing, The Wukong paper	Mid Project discussions and presentations start next week.
	Tue: Oct 17	Patient API, Incident Migration, AutoScale KV store	Project Discussions
	Thu: Oct 19	Harmful Post Detection, Distributed Fuzzy Inference	Project Discussions
	Tue: Oct 24	Incentivized Collaboration, Distributed Movie Recommendation, Model Splitting for Distributed ML	Project Discussions
12	Thu: Oct 26	Serverless Caching	InfiniCache, InfiniStore	Pocket, AWS Lambda	Coding Assignment 2: due Nov 5th
13	Tue: Oct 31	Serverless Storage + Parallel Processing	InfiniStore, Wukong	Hacker News, Review
14	Thu: Nov 2	Serverless Parallel Processing + Cloud Control Operations	Wukong, CNSBench	Wukong Docs
15	Tue: Nov 7	MapReduce Scheduling	SPARK, hatS	MapReduce Heterogeneity
16	Thu: Nov 9	Coding Assignment 2 design discussion + Demo + Cloud Resource Management I	Mesos, MOS	Mesos Video
17	Tue: Nov 14	Cloud Resource Management II	Google Borg, Borg: the next generation, Alibaba Trace Analysis, Borg to Kubernetes	Alibaba Microservice Trace Analysis, Borg Video	Coding Assignment 3: due Nov 20th
18	Thu: Nov 16	Deep Learning in Cloud	Horvod, Ray	Why Ray?, Spark Summit, Parameter Server
19	Tue: Nov 21	Deep Learning in Cloud II / Federated Learning as a Service	Alpa, IBM FL, TiFL	AlpaServe
	Thu: Nov 23	Thanksgiving Break
20	Tue: Nov 28	Communication in FL	FedAT	OORT
	Thu: Nov 30	Patient API,  Incident Migration,  Auto Scale KV Store	Project Presentations		Homework Assignment 3: due Today
	Tue: Dec 5	Harmful Post Detection, Distributed Fuzzy Inference, Incentivized Learning	Project Presentations
	Thu: Dec 7	Model Splitting for ML, Online Movie Rec.	Project Presentations
	Tue: Dec 12	Final Report Discussion

 

Final Project 

The research project can be done in a group of 2 to 3 students. The project will be judged on the following criteria:

• Proposal and literature review (10%) 
• Midterm presentation (10%) 
• Final presentation (10%) 
• Project Report (20%)
  
  

Project Proposal Report

Recommended structure for the Proposal Report is:

• Title + abstract.
• Related work: Need to have ~15 references. Reference could be papers, websites (e.g., GitHub repos, Internet articles, blog entries, etc.).
• Motivation + concrete problem statement (need not be formal, only concrete).
• A brief description of your expected approaches/system_design/experiment_design to solve the problem.
• Timeline/Expected milestones for achieving your goals.
  
  

Mid-way Project Presentation

You should present the current progress of your final project including design, related work review, any results if you have some (to show motivation or evaluation).

 

Final Project Presentation

You should present the final project including design, related work, and final results to showcase your project.

 

Project Final Report

The final deliverable of this course is a conference-quality paper. Focus on the following Questions 0 through 5 (an extended version of Question 0 through 3 in the Checkpoint Report requirement):

• Question 0: What is/are your main hypothesis/hypotheses in the project? This is a one-sentence summary of your paper. Position your paper w.r.t. other related work on the same or even similar problems (reuse relevant parts of your survey report here!). Show to the reader that your work is on a unique point in the design spectrum. Don’t overstate it, don’t understate it either. This statement is related to Question 5 below (yes, it’s a chicken and egg problem!)
• Question 1: What are your goals?
• Question 2: What are your techniques?
• Question 3: What do your techniques gain you?
• Question 4: What do your techniques lose you? That is, what are the tradeoffs.
• Question 5: What do your experiments tell you? (This is related to Question 0, 3, 4 above, and again, it’s an egg and chicken problem!) Show that your data supports your hypothesis, and show the gains, losses, and tradeoffs by using your techniques.

As well as pay attention to:

• Importance of problem,
• Novelty of solution,
• Evaluation of solution,
• Clarity of Presentation,
• Nits (grammar, references, etc.).

It is common that the implementation or experimental methodology gets adapted in the course of your investigation. And therefore, given that this is your final write-up, please make sure to refine/change/expand based on your mid-way presentation feedback to reflect your latest implementation & evaluation activities.

Leaderboard - KV performance

 

Group Members	Throughput 3 KV Stores	Average Latency	Ranking
Tejasvi Bansal, Rishabh Agarwal	7362	3.1 ms
Ammar Ahmed, Connor Howe	8240	2.87 ms
Xinran Wang, Jiaxiang Tang, Minrui Tian	7900	3.20 ms
Jon Meshesha, Andrew Owens	4456	3.53 ms
Pat Johnson	9572	3.1 ms
Leo Dong, Michael Andrev, Apekshik	30332	1.43 ms	2nd
Sahil Raina, Jingxian Chai, Shunichi Sawamura	25588	2.9 ms	3rd
Aahan Tyagi, Hemanth Kumar, Vishal Kancharla	83865	3.2 ms	1st