Live data from Hacker News

Mysteries of Telegram Data Centers (2022)

dev.moe

1–10 of 166 posts

Re: Mysteries of Telegram Data Centers (2022)

#5

This strikes me as a huge amount of custom code and technical debt. Every new software dev probably has to learn this. Why not a sticky master election per user, and have no special data centers?

From what I have read, they only have ~30 employees. They're not exactly onboarding a lot of new people here.

Re: Mysteries of Telegram Data Centers (2022)

#8

i'm far from an authority on content delivery or whatever, but the first thing I thought of was what a bizarre way to setup your infrastructure!

idk, they probably tried to get people on DC's as close to their location as possible. Using your phone number's country code might seem like a good way to do this at first, and they probably didn't give it much more thought before building the whole thing on this idea.

Re: Mysteries of Telegram Data Centers (2022)

#9

This strikes me as a huge amount of custom code and technical debt. Every new software dev probably has to learn this. Why not a sticky master election per user, and have no special data centers?

It makes sense: European users are assigned to EU data center, and Chinese to the one closer to them. The "custom code" should not be complicated, just a map of country to DC.

You are suggesting to develop a compicated solution (spend money) when current simple one is working ok without any elections.

Re: Mysteries of Telegram Data Centers (2022)

#10
DC2 is the first connection point of all MTProto clients.

Any DC may refuse a request and force the client to switch DC.

Profile URL doesn't show where messages/chats/channels are stored, as telegram has two dedicated DCs mostly for media. The rest DCs allow media with bandwidth being throttled.

Post reply on HN