Skip to content

Latest commit

 

History

History
107 lines (71 loc) · 2.78 KB

discussion.md

File metadata and controls

107 lines (71 loc) · 2.78 KB

Identify the Entities and Relations

Entities

  • User: name, username, password, phone
  • Expense: description, date
  • Group: name

Relationships

  • User m--- participates in ---n Expenses
  • User m--- member of ---n Group
  • User 1--- admin of ---n Group
  • Expense n--- belongs to ---0/1 Group

DB Schema?

  • How important is "how you're persisting the data"?
  • this is a detail - talk about this later

Services / Interfaces

Other topics to explore

  1. Domain Driven Design
  2. Clean Architecture

Limits on groups

  • Why does Whatsapp have a limit on number of group members
  • Why does Facebook allow max 3000 friends
  • Why does Linkedin allow max 30,000 connections

Choosing b/w SQL vs NoSQL

  1. Your default choice should be relational databases - SQL
  • Relational databases are a mature technology - 1970
  • string ACID guarantees
  • they're fast enough for most purposes.
  • unless you're Google / unless you're dealing with billions of users
  • 10s of millions of users / 100 million users - SQL is good enough!
  • SQL is slow -> myth
  • SQL is slow if nd only if you're dealing with big data, or a lot of concurrency
  1. Whenever you choose NoSQL, you have to justify it!
  • Recent technology
  • A lot of misconceptions around these technologies
  • If you commit something to Mongodb, it does NOT actually get written immideately
  • It is in the memory, which might crash!

When would you use a document style DB - MongoDB / CouchDB / CouchBase ?

  • you don't care about consistency!
  • 1% of the likes fail to be persisted, is that a problem?
  • 1% of your money transactions get lost, is that a problem!?
  • when your data is unstructured / semistructured
  • or your team is moving very fast - agile - constantly changing the db schema

When would you use a wide-column database - Cassandra

  • data is stored column-wise rather than row-wise
  • in most analytics applications, you want to run a query on a column
  • get the average age of users
  • get the average rating by users

Cloud dbs

aws amplify PouchDB (sits on top of Couchdb) Rxdb (sits of top of Pouchdb)

  • are reactive
  • are offline first
  • they sync with the cloud service

S3 storage - blob/file storage Hbase

At the top level - CTO / Tech head - you will have to decide the tech stack

  • the more you know the better!

NoSQL database - does NOT lack relations Mongodb - have the id of another document in a document

Relations are very important to you, and you're doing nested queries - social network GraphDB - neo4j

CAP theorem

SQL dbs also have a lot of tooling and a huge community MongoDB -> tomorrow, you run into a problem -> post it in stackoverflow and noone replies db is normalized, and joins are costly - add a redis based cache layer to your db -> do the denormalization

Group 0/1 ---- * Expense

pragy@scaler.com