[numair]

Oh man, where do we even begin. $1.5B would be a great number if it's true. The fact that the commenter compares that amount to the $300M they claim Netflix spends each year shows that they ... are ... a lot less intelligent than they probably think they are.

Twitter: Billions of little updates that invalidate billions of cached data sets, every single day, that need to be analyzed and delivered very quickly.

Netflix: Maybe millions of static files that do not ever change, that need to be shuttled to the closest CDN peer and cross referenced against profile and billing data before allowing access.

How does anyone even begin to compare these two engineering tasks and think they have anything in common? ... And, looking at their timeline, they then went on to compare Twitter's infra to WhatsApp in the 2010s, when it did nothing other than store-and-forward over an Erlang bus! And thinks this infra has something in common with what it takes to run Twitter! What!

I guess it shouldn't surprise me that a false expert with such a mediocre grasp of the realities of social infrastructure would be a major Elon fanboy.

[Tanjreeve]

The whole twitter/Elon thing it's been very noticeable that a lot of the mindset divide was people who deal with software and the guts of distributed systems and performance Vs "AI researchers"/tech bloggers people who think once youve got your AI algorithm sorted in pandas or your "twitter hello world" UI you just lob it over the fence and it's done automatically.

[jmoak3]

Always remind them of their S.I.N.S. (Scale is not simple) haha

[badwolf]

The common trope lately is to compare Twitter now to Whatsapp back in the days. It's so disingenuous and tiring.

[Tanjreeve]

The same people doing that now would've probably looked at WhatsApp and been coming out with trivialised comparisons and saying Erlang is some hipster language etc and why arent they sharding more or using Hadoop because "big data" etc.

[Retric]

Netflix also gets free hoisting inside most ISP’s which really changes these calculations.

[Kye]

Netflix probably re-encodes for new devices (to avoid delays doing it on the fly) regularly like Google does for YouTube. That compute time might make up some of the difference despite the different needs.