Hacker News Reader

Comment by toomuchtodo | Hacker News Reader

> They want good engineers to solve challenging problems.

Serve bits. Suggest movies based on coarse signals. Is there much left?

Netflix is not SpaceX nor Tesla; its an online Blockbuster Video. Serving video in 2017 is not complicated.

I work on "serving bits". It is not quite a simple as it sounds at our scale. I've specifically worked on optimizing our FreeBSD kernel, along with several other engineers on my team. We're at point where we can serve over 90Gb/s of TLS encrypted traffic (many tens of thousands of TCP streams) from a single-socket 16 core Xeon, without any crypto offloads.

There are a lot of challenging problems in any industry once you dig deep enough. There are tons of brilliant people around me, doing super cool things that I can barely fathom (like the recommendation system). Any one of those things can be trivialized, but each is somebody's pride and joy, and a marvel of design and engineering.

Definitely appreciate you taking the time to respond to my comment. Netflix makes it look easy, so from my outside perspective, its easy to dismiss the problems as "solved".

You could say the same thing about chat apps (What's App) and picture sharing (Instagram)...

>Is there much left?

You might want to look at the Netflix tech blog - lots of interesting posts covering all kinds of topics.

> Serve bits. Suggest movies based on coarse signals. Is there much left?

I didn't want to answer your question right away since it felt like you were trolling at first. But here's my answer. There's indeed, quite a bit more than just serving bits and title suggestions.

The "left over" portion includes things such as getting those bits to move effortlessly to the multitude of TV's, set-top boxes, and streaming sticks, not to mention the effort and engineering to get those movies and shows streaming effortlessly to your mobile devices.

We attempt to optimize every part of the user experience so you can get to watching your shows immediately regardless of what device you choose to pick up.

I just read through the OP's other post about PMCs, and I would beg to differ: http://www.brendangregg.com/blog/2017-05-04/the-pmcs-of-ec2....

Supplying sufficient video to be responsible for 25% of consumer internet bandwidth is a different story.

Can you be more specific? Do you mean producing their original content? Or rendering renditions for each bitrate, and then pushing those renditions out to their custom CDN appliances? (OpenConnect)

I can most definitely get behind original content production being hard. But pushing a couple hundred TB (or even low double digit PB) of data out to CDNs doesn't require a fleet of devs.

Total content is likely closer to 100 PB. The "25% of consumer bandwidth" is a proxy for how heavily Netflix is used.

Take a look at some of their open source projects and their tech blog to get a feel for the problems they solve at their scale.

https://netflix.github.io/ https://medium.com/netflix-techblog

Some things I find personally interesting:

They run a 5000 node Cassandra cluster.

They can failover an entire datacenter in 27 minutes.

Akamai can do that for you just fine.

Apparently, Akamai does about 30 Terabits/s[0]. Netflix does "tens of terabits per second"[1].

Essentially, Netflix is about as big as Akamai when it comes to content delivery.

[0] https://www.akamai.com/us/en/about/facts-figures.jsp

[1] https://media.netflix.com/en/company-blog/how-netflix-works-...

This is such a naive, ignorant statement that no wonder you would never get a job at Netflix.

Reply on news.ycombinator.com