78 points by jonashoechst 128 days ago
Very interesting project, I also suggest to take a look the authors current research efforts. Essentially he designed a language to program MP-TCP schedulers, making it straightforward to generate schedulers for diverse use cases such as low latency or high bandwidth applications.
TL;DR: can you connect to this website with an IETF experimental standard that hasnt been upstreamed into the linux kernel and only started using function pointers in 2015?
Be that as it may, it's available to apps in ios 11, so it seems likely to get more traction.
interestingly enough this is largely due to the ios subsystems use of MPTCP for siri. It also shows up in F5 products, albeit one must wonder if its ever advertised for use in genuine load balancing
So what's the current status of the upstreaming work? It sounds like there's a good start discussed in the paper, in removing a lot of conditional and duplicated code and using some more cleanly separated layers, but that was two years ago, what's happened since?
We are back online (at least a bit ;-) ... or a byte?
Part I found most interesting:
>Visitor counter: 63684 in total, 10 with MPTCP since October the 13th, 2017.
A few of those are from me after I installed mptcp, so the counter seems to be working.
How will MPTCP effect the efficiency of RF mesh network architectures?
So, for example if I have 3G and Wi-Fi connections what will be my IP from the server's perspective?
MPTCP's bandwdith aggregation properties are cool---you can effectively use all your network interfaces for (MP)TCP connections.
Has anyone deployed it in data-center/cloud environments? I know that both Apple and Android support it for the mobile devices to use WiFi and LTE at the same time.
If you have multiple ISPs in your datacenters, you can announce different IP blocks over each upstream at each site and configure full mesh MPTCP between your sites to expose as much path diversity as possible, and have MPTCP switch to the most reliable path(s), kind of like a poor man's MPLS (in my experience bulk internet bandwidth is significantly cheaper than MPLS). I've implemented this as a demo not carrying a ton of traffic and it seems to work really well, but never got around to doing it full scale. Would love to do so.
I know there are Android builds you can install with mptcp support, but I didn't think it had made it into mainstream android phones, do you have a link?
Google doesn't support MPTCP.
They think QUIC streams (supported by Chrome and google services) serve the same purpose better.
Within a datacenter, you'd be more likely to use LACP and/or ECMP.
Though, with many LACP hash implementations, you end up de facto restricted to a single link per TCP flow.
ECMP works great with MPTCP :-)
Best used for cross-DC traffic over public networks if you have access to multiple physical links, and no transparent DC-Internet routing
Does anyone have any success with macOS High Sierra and Safari? I would've expected it to work since iOS 11 reportedly supports it but so far I haven't had any success with my Mac despite trying different networks.
The iOS app developer has to explicitly turn on MPTCP, e.g., like shown here https://github.com/below/MultipathTCP
Very nice, but the website looks down. Looking forward to play more.
Ah, you are evidently using zero-path TCP.