1) Reducing the unmanaged buffers on the devices I have (all Linux-based) and observing the effects on latency and throughput. This involves a bit of kernel hacking. As kernel hackers go, I’m a little unusual in that usually I work on rt based kernels in the first place, because I do audio editing and production. At the moment, I’m merely using the 2.6.37 kernel as a base, as it’s common between all my devices and the rt patches have frozen at .33.
2) High on my personal list is exploring the inter-relationship between the mq qdisc and the wireless device drivers themselves. Buffering is ok. Unmanaged buffering is not. Wireless has special problems for which most people have been papering over with excessive buffering and retransmit settings, especially on newer devices.
3) Also related to that is playing with the SFB qdisc which seems to have a lot of potential to manage flows at the gateway and laptop level.
4) In the context of real-time, I’d like to be exploring the interaction of HTB and device driver buffering in the context of having smaller quantums for the kernel and device driver(s) to be making their decisions.
5) I’m also helping maintain and upgrade the bufferbloat website, and keep the public conversation(s) on track, and recruiting helpers which is eating far more of my time than I’d like.
6) Writing articles on describing the mis-understood portions of the IP stack.
I can be found in the #bufferbloat channel(s) on irc.
Save the Mice Diffserv statistics
Improving DSCP support in Linux
There’s a rule of thumb for the upper bound in two SIGCOMM-award
papers; I generally use the simple one from the Mathis paper.
bandwidth = (MSS / RTT) * (C / sqrt(loss))
MSS = maximum segment size
RTT = round trip time
C = a constant between 0.87 and 1.31, depending on the ACK strategy
and type of loss
sqrt(loss) = square root of the probability of losing a packet
So, for a given connection, your expected throughput scales linearly
with the size of your packets, and inversely with the RTT and
I have an unverified/untested Python implementation of the equations
I wonder how jumbo frames play into bufferbloat. When queues are in
terms of number of packets (not bytes), jumbo frames make the problem
Subject: Re: Understanding Bandwidth-Delay Product in Mobile Ad Hoc
On 01/18/2011 11:27 AM, Dave Täht wrote:
> Don’t know if anyone sent this one to you yet.
> Also, PLUGFEST!()! Sounds like a good spot to do some testing in
No, even better is to educate the UNH Interoperability lab, and have
them test everything.
Dunno if they are competent to test 802.11 full up or not for what we
need; but the task of educating them will make that clear.
See: http://www.iol.unh.edu/ - Jim