[FWD] Pi in the sky: Calculating a record-breaking 31.4 trillion digits of Archimedes’ constant on Google Cloud

Notes:

• Google Blog: [blog]
• Method: Chudnovsky formula: [wiki]
• L. Milla: A Detailed Proof of the Chudnovsky Formula: [paper]

[Books] Programming Pearls (2nd Edition)

Bought a Chinese translation of Programming Pearls (2nd Edition).



Super recommend to read!



Quote from this book:

Q: It seems that most columns emphasize the design process. Can you summarize your advice on that topic?

• Work on the right problem.
• Explore the design space of solutions.
• Look at the data.
• Use the back of the envelope.
• Exploit symmetry.
• Design with components.
• Build prototypes.
• Make tradeoffs when you have to.
• Keep it simple.
• Strive for elegance.

Also check this for designing machine learning systems: Hidden Technical Debt in Machine Learning Systems [paper]. 

[FWD] Cost of Living Comparison

Example: Cost of Living Comparison Between Taipei and Singapore

https://www.numbeo.com/cost-of-living/compare_cities.jsp?country1=Taiwan&city1=Taipei&country2=Singapore&city2=Singapore&amount=1%2C000%2C000.00&displayCurrency=TWD

+30% but we can explain the whole details to clarify the correctness of these metrics.

Notes on System Design

Warning: All materials are prepared for Facebook onsite interview but I failed.


Case Study:

• Netflix (mysterious)
• Learning a Personalized Homepage – Netflix TechBlog – Medium (2015): Article
• The Netflix Recommender System: Algorithms, Business Value, and Innovation (2016): Paper
• Benchmarking Cassandra Scalability on AWS — Over a million writes per second (2011): Article
• A Multi-Armed Bandit Framework for Recommendations at Netflix | DataEngConf SF '18 (2018): YouTube
• Google
• Building Software Systems at Google and Lessons Learned (2010): Slide | YouTube
• Google Production Environment (2018): YouTube
• The Google File System - Research (2003): Paper
• MapReduce: Simplified Data Processing on Large Clusters (2004): Paper
• Bigtable: A Distributed Storage System for Structured Data (2006): Paper
• Differential Synchronization (2009): Paper | YouTube
• The Paxos Algorithm (2018): YouTube
• Yahoo
• The Hadoop Distributed File System (2010): Paper
• Analyzing Google File System and Hadoop Distributed File System (2016): Article 
• Youtube
• Deep Neural Networks for YouTube Recommendations (2016): Paper
• RecSys 2016: Paper Session 6 - Deep Neural Networks for YouTube Recommendations (2016): YouTube
• Facebook
• Balancing Multi-Tenancy and Isolation at 4 Billion QPS (2015): YouTube
• TAO: Facebook's Distributed Data Store for the Social Graph - Usenix (2013): Paper
• Finding a needle in Haystack: Facebook's photo storage - Usenix (2010): Paper
• Cassandra - A Decentralized Structured Storage System (2009): Paper
• Instagram
• Scaling Instagram (QCon London 2017): YouTube
• Amazon
• Dynamo: Amazon's Highly Available Key-value Store (2007): Paper
• Slack
• How Slack Works (QCon San Francisco 2016): YouTube
• Uber
• Project Mezzanine: The Great Migration (2015): Article
• Designing Schemaless, Uber Engineering’s Scalable Datastore Using MySQL (2016): Article Part 1 & Part2 & Part3
• The Uber Engineering Tech Stack, Part I: The Foundation (2016): Article
• The Uber Engineering Tech Stack, Part II: The Edge and Beyond (2016): Article
• How Uber Scales Their Real-Time Market Platform (2015): Article

Interview Practices:

• URL shortener system design | tinyurl system design | bitly system design (2018): YouTube
• NETFLIX System design | software architecture for netflix (2018): YouTube
• UBER System design | OLA system design | uber architecture | amazon interview question (2018): YouTube
• Twitter system design | twitter Software architecture | twitter interview questions (2018): YouTube
• Whatsapp System design or software architecture (2018): YouTube

Notes on Coding Interviews

• Search on YouTube.
• Practice. Practice. Practice.
• Practice a lot of problems? Not so effective.
• Effective way: build up your knowledge from Leetcode.
• Organize classical problems into several categories.
• Excellent example: https://cspiration.com/leetcodeClassification#103.
• My 5 categories: (All-in-one)[doc]
• Foundations: [doc]
• Array
• String
• Math
• Data Structure: [doc]
• Stack
• PriorityQueue
• LinkedList
• Binary Search Tree
• Advanced Design: [doc]
• Backtracking
• Dynamic Programming (too many dissimilar problems, and hard to prepare)
• Union Find
• Trie
• Graph Algorithms: [doc]
• Graph
• Tree
• DFS & BFS
• Topological Sort
• Selected Topics: [doc]
• System Design
• Random
• Matrix
• Bit Manipulation
• Computational Geometry
• Read articles in the Discuss.
• Identify all possible solutions
• Different algorithms (from brute force to effective algorithms).
• Same algorithm with different implementations (e.g., segment tree).
• (My documents are not completed. Try to build up your own documents.)
• Could analyze time complexity and space complexity of each solution.
• Analyze line by line.
• Understand all details of build-in libraries.
• Know which solution is the best under certain assumption.
• Official solutions are not always concise (and the best). But it must be correct.
• Improve your code review ability.
• Read codes in the Submission (if possible).
• Not always correct.
• Learn from undocumented solutions. Need more time to understand.
• Write down your concise code in your style.
• About 20 ~ 40 lines in Java.
• Concise but not tricky. Tricky codes are not easy to remember.
• Possible to reproduce in the real interview.
• Lengthy codes might be regarded as messy codes. 
• Just copy or imitate other's code. 
• Not shame.
• Save your time.
• Align to your style (e.g., binary search, naming convention and so on).
• Practice makes perfect.
• Identify your weakness.
• Could prove the correctness of your code.
• Demonstrate your codes line by line.
• Design meaningful test cases.