Hacker News new | past | comments | ask | show | jobs | submit login

Can I ask something that will sound incredibly naive?

How complicated needs to be an IT infrastructure that lets you book a taxi via an app and pay the driver? Is it really something to throw billions at each year? Sometimes I feel these companies are throwing billions in IT simply because they're expected to be disruptive technology companies, and not for real technical reasons.

I imagine a driver-passenger matchmaking system could be craigslist level of simplicity, but Uber and Lyft want to suck up every piece of data: ingesting and analyzing where people are when they open the app, how many people give up on finding a driver, how long every driver is idling on the side of the road, whether a user is probably a cop, etc etc

They use this data to update their surge pricing by the minute for all these different geofences all over the world... yeah, I think having the self image of a technology company that wanted to attract the best talent definitely created a lot of expensive problems for itself.

At a minimum drivers need to be tracked in real time from the point they respond to the point passengers are picked up, and preferably tracked throughout trips due to various concerns, so it definitely can’t be Craigslist level of simplicity.

There are similar, smaller services (like Via, or myTaxi in Europe) that work fine and have nowhere near the money that Uber spends on their tech stack.

Ride sharing is extremely local, so it's not like you need any cleverly scalable tech. Most places will be tracking a few hundred cars at any given moment. Only in very large cities like NYC would you need to spend a little more on big servers.

What Uber presumably does is collect and analyze a lot of real-time data, which isn't strictly necessary to provide the service they're providing.

Uber as a product is very technologically complex, mainly because of the scale it operates at. You can read about some of their tech at https://eng.uber.com.

From that page I can see that they created their own deep learning framework, their own AI conversational agents, their own data science platform, their own geospatial indexing system, their own web applications framework, etc. The list seems to go on and on and on.

Essentially they seem to be repeating whatever Google and Facebook are doing: but those are pure technology companies that serve two orders of magnitude more customers and have contributed building the web as we know it. Uber is a taxi company.

Only if you want to have a huge and complex system. If you have small and localized ones, it should be very cheap.

Ye, like one VM running per city or region. I feel they are just sinking money into software and administration of software.

WhatsApp has shown that you can actually build large-scale system infrastructures very efficiently. And they had global service, Uber could profit from sharding as you suggest.

Apply for a job at Uber proposing a cut in engineering expenses with your plan and become instantly rich.

And never get hired because the people doing interviews have to protect their jobs :)

> Uber as a product is very technologically complex, mainly because of the scale it operates at.

Also because when Uber was young, it was not motivated to seek profits or consider expenses, it was just growth at any cost. Once that work culture is established, it can be hard to shake.

It doesn’t!

You can read here on HN how their large IT staff spends most of its time migrating back and forth between MySQL and Postgres.

Applications are open for YC Winter 2020

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact