News for hackers. Stuff that matters.
Re:Pirate Bay founder thinks Parler’s inability to stay onli... by toast0 on Thursday January 14, 2021 @03:11AM (#25773552)
Ideally, for redundancy, you'd run in more than one datacenter. That's hard to do with only one /24; most networks won't take BGP advertisements more specific than /24, so you can't easily run half your ips at each DC. Maybe HE allows it for transit customers, but it makes the routing messier.

They could be doing anycast, but that's pretty sophisticated. They could also only advertise the /24 from one DC at a time, but that's error prone. Or maybe advertise from both and forward to the live DC over a private link/VPN.

I do see in Octoberish, they were peering with a second network, and they have second network in their IRR records. Could just be so they can use one of the DDoS filter companies that advertises your netblock, and sends you clean traffic. Can't do that (easily) if you get IPs from your transit provider and it's not a service they support.

Re:Pirate Bay founder thinks Parler’s inability to stay onli... by treesknees on Saturday January 16, 2021 @02:19PM (#25804625)
You can still have redundancy. Basically active + backup config, where the other datacenter takes over and announces the /24 if the active goes down. They could certainly be utilizing "oob" IP connectivity to coordinate failover instead of using their own IP space.