Live data from Hacker News

“Multipath” TCP Could Turbocharge Bandwidth

technologyreview.com

11–20 of 30 posts

Re: “Multipath” TCP Could Turbocharge Bandwidth

#11
Multiparty sounds great and all. But I'd be happy if they could make a laptop use both a hard-wired Ethernet and a wifi connection at the same time. Is it really that difficult?

Not for bandwidth, I work with embedded systems. If the wifi is active, then it's not possible to connect to an IP address on the Ethernet port (OS X and Win7). Subnet doesn't matter, and re-arranging the priorities of the network devices doesn't help either.

Re: “Multipath” TCP Could Turbocharge Bandwidth

#12

Multiparty sounds great and all. But I'd be happy if they could make a laptop use both a hard-wired Ethernet and a wifi connection at the same time. Is it really that difficult? Not for bandwidth, I work with embedded systems. If the wifi is active, then it's not possible to connect to an IP address on the Ethernet port (OS X and Win7). Subnet doesn't matter, and re-arranging the priorities of the network devices doe…

You should read up on how routing works. In particular - source-based routing.

Re: “Multipath” TCP Could Turbocharge Bandwidth

#13

Multiparty sounds great and all. But I'd be happy if they could make a laptop use both a hard-wired Ethernet and a wifi connection at the same time. Is it really that difficult? Not for bandwidth, I work with embedded systems. If the wifi is active, then it's not possible to connect to an IP address on the Ethernet port (OS X and Win7). Subnet doesn't matter, and re-arranging the priorities of the network devices doe…

It may just be my laptop, but I was able to get both WiFi and Ethernet working on an HP Elitebook 8460p after toggling a setting in the BIOS.

Re: “Multipath” TCP Could Turbocharge Bandwidth

#14
post #3

Don't both endpoints have to support Multipath TCP? So I imagine we wouldn't see real gains for standard HTTP traffic in a good long while. Unless Apple passes the traffic through some intermediate proxy, of course...

Yes, as the article suggests, Apple is using it for reliability and latency for their built-in services. But if they can do it, we can't be that far from NetFlix, Google and others being able to flip the switch for their high bandwidth products. I am curious whether iOS exposes multipath to applications.

I imagine it'd be pretty similar to the adoption curve SPDY has seen. First Chrome implemented it to talk to Google's servers, then a few other websites, and now nginx will do it for you "for free."

Re: “Multipath” TCP Could Turbocharge Bandwidth

#15
Aren't well-seeded torrents essentially multipath TCP? The main thing they get around seems to be per-flow limits instituted on network provider routers and SPOFfy fileservers, the first being a flow deSPOF, and the second being an endpoint deSPOF. Multipath TCP (though I'm not very familiar with it) seems to be only a flow deSPOF, and therefore only solves half the resiliency domains solved by torrents. Of course, they are no doubt faster to set up and don't require half the world to be interested in something .. both significant benefits.

Re: “Multipath” TCP Could Turbocharge Bandwidth

#16

Multiparty sounds great and all. But I'd be happy if they could make a laptop use both a hard-wired Ethernet and a wifi connection at the same time. Is it really that difficult? Not for bandwidth, I work with embedded systems. If the wifi is active, then it's not possible to connect to an IP address on the Ethernet port (OS X and Win7). Subnet doesn't matter, and re-arranging the priorities of the network devices doe…

You can "shotgun" connections but hardwired and wifi usually isn't a win since the wifi will be more error prone and higher latency than the wired. So if you have both you should probably just do the wired.

Re: “Multipath” TCP Could Turbocharge Bandwidth

#17

Don't both endpoints have to support Multipath TCP? So I imagine we wouldn't see real gains for standard HTTP traffic in a good long while. Unless Apple passes the traffic through some intermediate proxy, of course...

Multipath TCP doesn't magically increase the amount of bandwidth either, it just allows a single TCP-like connection to use multiple internet connections. It would allow you to use all bandwidth available on multiple connections in a single TCP connection but right now we can already use all bandwidth available over multiple separate TCP connections.

Re: “Multipath” TCP Could Turbocharge Bandwidth

#18

I've had this discussion about a dozen times now. MPTCP is a hack, plain and simple. It does not address the fundamental flaws of TCP, nor does it attempt to use something better suited for the job of, say, streaming media like SCTP. TCP is broken in today's age. It's still reliant on upstream single path routes that are mainly devised via BGP and OSPF. These two routing protocols are showing age and suck at multi pa…

MPTCP does not claim to address the fundamental flaws of TCP. It allows a single TCP connection to be transmitted accross multiple TCP connections on different interfaces. No more, no less. If that's a hack for you, then a hack it is.

Re: “Multipath” TCP Could Turbocharge Bandwidth

#20

Multiparty sounds great and all. But I'd be happy if they could make a laptop use both a hard-wired Ethernet and a wifi connection at the same time. Is it really that difficult? Not for bandwidth, I work with embedded systems. If the wifi is active, then it's not possible to connect to an IP address on the Ethernet port (OS X and Win7). Subnet doesn't matter, and re-arranging the priorities of the network devices doe…

My laptop can use both connection at the same time. Maybe you need a better OS.
Post reply on HN