Hacker News Reader

Comment by 0xy | Hacker News Reader

Why was it trivial for FB to scale WhatsApp from 450,000,000 users at acquisition to over 3,300,000,000 now?

Seems like an insane achievement to scale a messaging service to half of the world.

8x scaling over a time period where the available size and performance of single-image systems increased by more than 8x seems… if not trivial, at least not an “insane achievement.”

Scaling a product like that is an incredible engineering achievement, but I am speaking specifically about product development. What other products has Facebook innovated internally, worked through the product-market fit, and scaled itself? Again, my point is that they have a ton of resources and a huge existing user base, but the only successes they seem to have are purchasing companies that are ramping. They deploy engineers and connect to the existing user base, but they aren't innovative in product design.

facebook itself is a myspace or nexopia clone

the targetting pixel and like button are the most obvious innovations?

facebook is a tool for building products on, rather than a product in and of itself.

Was it just because WhatsApp was already generally well designed when they purchased it?

I think that might be common enough these days through sound horizontal scaling principals and especially services that scale for you, but Whatsapp launched in 2009, pretty much the turning point for the tech that enables "web scale".

At that time AWS was just rolling out ELB and RDS, people were still fulfilling their (and most) roles on EC2 servers or even more likely dedicated servers / VPS that took day(s) to commission and might have even been setup by hand, there was no Docker, GitHub was new and Actions, Jenkins etc were years away, and there were very few PaaS- or IaaS-type offerings IIRC just a very nascent Heroku and Google App Engine.

WhatsApp was built on Erlang, though, which was built to scale in the 80s.

ejabberd right?

“Scaling” as in “making sure the infrastructure can handle much higher load”, or as in “making sure the product remains genuinely useful to people, so that the user numbers go up and not down”? For both, it didn’t happen by itself, but it’s far from rocket science. A sane team of 15-20 people can do it.

Scaling a product from 0 to 450 MM is a humongous achievement.

Going from 450 to 3300 seems rather small in terms of technical achievement. If you can serve 450 you can serve 3300 MM frankly.

Complete nonsense, as anyone who has worked on a service with scale would know. WhatsApp is not even close to the same architecture as it was at the time of acquisition.

Not to mention the load constraints did not grow 7x, it's orders of magnitude more load increase than that.

As the world's mobile data speeds have skyrocketed in speeds and plummeted in price, WhatsApp is now handling orders of magnitude more data per user than when they served merely 450. And it's all complicated rich media which was not handled at the time.

WhatsApp's file limit at the time of acquisition was 16 MB. Today, it's 2 GB. Per file. It didn't have voice or video calling at acquisition either, also highly complex and expensive to run.

WhatsApp grew 7x in user base, but the scale of data and technical complexity grew exponentially more.

If this were so easy, history would not be littered with high profile scaling failures from messaging apps. Most notably Signal, which couldn't handle a 1.5x influx of users from WhatsApp and was down for weeks.

Claiming a 7x scale is technically easy while your data-per-user growth grows probably 50% a year or more is patently ridiculous.

Reply on news.ycombinator.com