Untitled.md

<< ---------------------------------------------------------------- >>

--- Last Modified: $= dv.current().file.mtime

Uber

<< ---------------------------------------------------------------- >> https://www.youtube.com/watch?v=giwa8Hc0niY&list=PLjTveVh7FakJOoY6GPZGWHHl4shhDT8iV&index=11

Functional Requirements

1. request a ride between two locations
2. see drivers in your area upon request, drivers can see you
3. Provide an expected amount of distance and time the ride will take
   1. in this one we will abstract it to google maps
4. compute the exact route that you are to take to get to the destination

Computational: assume 1 billion users take a 20 minute weekly ride gets to aprox. 20 million rides at a time assuming 64 bit lat/long and 64 user id → 1 gb of info 20 million/ 20 minute = 15000 requests per minute to the DB

Do have to use geo hashing and geo sharding(have to monitor it and twinker it a lot, a lot of 3rd party libraries to help with this)

Have to deal with race conditions when approaching drivers to accept a ride request.

You can just do a predicate lock for a rides info table. this will be slow since the predicate it self is a query that has to search the table for a given ride ID, so to make it optimal you have to index it using the ride ID.

You can also just create the ride id with an empty driver row in the table and lock it when contacting a driver.