“Multipath” TCP Could Turbocharge Bandwidth

Re:“Multipath” TCP Could Turbocharge Bandwidth by zero_intp on Tuesday October 01, 2013 @06:12AM (#6475194)

Nitpicking: ISIS is /FAR/ more prevalent than OSPF in the really real world of ISPs. Both run Dijkstra at their core, but ISIS handles multipath much more cleanly.

Bloat is a very interesting conversation, most providers are not doing sufficient analysis of the interface buffer size in relation to latency. It would be very neat to see Juniper/Cisco building interface level auto-queue management to reduce the latency inducing over buffering that occurs on badly tweaked backbone links.

[ Parent | on HN ]

Re:“Multipath” TCP Could Turbocharge Bandwidth by windexh8er on Tuesday October 01, 2013 @10:21AM (#6476321)

Maybe in the EU - but ISIS fell off the bandwagon in the US years ago. I'd be curious to your claim, I've rarely run into ISIS in prod and I've had my hands on a lot of G2000 networks.

Cisco and Juniper are dinosaurs, the real disruption is being done by those smaller orgs not focused on pushing overpriced big-iron.

edit: I am a network engineer by trade, and have worked in large ISPs over the years.

[ Parent | on HN ]

Re:“Multipath” TCP Could Turbocharge Bandwidth by zero_intp on Wednesday October 02, 2013 @02:34PM (#6484329)

All of the backbone providers use ISIS. OSPF does not scale as well (# of vertices).

I am the IGP principle at a Tier 1.

[ Parent | on HN ]