Tired of relying on Big Tech to enable collaboration, peer-to-peer enthusiasts are creating a new model that cuts out the middleman. (That’s you, Google.)
Really boring article. It takes like 5 pages to get to what it’s tryna say: use CRDT, which is a real time collaborative editing algo, and CRDT is awesome. Problem is internet topologies and all the weird stuff that goes on
Take webrtc, which is exactly meant for p2p data streaming between arbitrary peers. Well, the dirty secret is that for it to work well, you usually need a TURN server, which is a centralized data relay. Unless you’re connecting over local network, the turn server serves to literally send all the webrtc data to the peers
When you think about this some more, many many apps would be worse p2p. Think about anything with a CDN delivering video or files. Obviously you want those videos pre-replicated worldwide so that they can be served asap. Ok so p2p is only for collaboration, fine
Next problem, there’s a good reason we all chose cloud. Even huge corps realized it would save them a ton of money to switch from their expensive private datacenters and staff. They were already paying money to some bomb shelter style server host, now they are just doing it virtually. And your engineers no longer have to drive out to wipe drives or replug wires, it’s all perfectly managed
Next problem, there’s a good reason we all chose cloud. Even huge corps realized it would save them a ton of money to switch from their expensive private datacenters and staff. They were already paying money to some bomb shelter style server host, now they are just doing it virtually. And your engineers no longer have to drive out to wipe drives or replug wires, it’s all perfectly managed
This part is just not true. Many companies are moving things back in house because of the cloud costs, along with how poorly the cloud actually turns out to be managed (at least the Microsoft one that most companies used for e-mail and collaboration). And the cloud never got easy enough to not need specialized employees, and in many cases, they’re more expensive than “on prem” employees were because it was the hot new buzzword for a while.
I can go into lots of technical details, but it’s worth pointing out that many huge corps are doing hybrid and using the cloud strictly for burst usage because the constant state costs are way way way cheaper if you own the servers. Which kind of makes sense - if you need a car for 2 days a year, you rent, but if you use it for hours a day, you buy.
Really boring article. It takes like 5 pages to get to what it’s tryna say: use CRDT, which is a real time collaborative editing algo, and CRDT is awesome. Problem is internet topologies and all the weird stuff that goes on
Take webrtc, which is exactly meant for p2p data streaming between arbitrary peers. Well, the dirty secret is that for it to work well, you usually need a TURN server, which is a centralized data relay. Unless you’re connecting over local network, the turn server serves to literally send all the webrtc data to the peers
When you think about this some more, many many apps would be worse p2p. Think about anything with a CDN delivering video or files. Obviously you want those videos pre-replicated worldwide so that they can be served asap. Ok so p2p is only for collaboration, fine
Next problem, there’s a good reason we all chose cloud. Even huge corps realized it would save them a ton of money to switch from their expensive private datacenters and staff. They were already paying money to some bomb shelter style server host, now they are just doing it virtually. And your engineers no longer have to drive out to wipe drives or replug wires, it’s all perfectly managed
This part is just not true. Many companies are moving things back in house because of the cloud costs, along with how poorly the cloud actually turns out to be managed (at least the Microsoft one that most companies used for e-mail and collaboration). And the cloud never got easy enough to not need specialized employees, and in many cases, they’re more expensive than “on prem” employees were because it was the hot new buzzword for a while.
I can go into lots of technical details, but it’s worth pointing out that many huge corps are doing hybrid and using the cloud strictly for burst usage because the constant state costs are way way way cheaper if you own the servers. Which kind of makes sense - if you need a car for 2 days a year, you rent, but if you use it for hours a day, you buy.
As a sys engineer looking forward to my cloud engineer pay increase I can confirm.