Comments on: Fundamental Progress Solving Bufferbloat

By: John McCabe-Dansted

John McCabe-Dansted — Sun, 13 May 2012 13:05:30 +0000

I guess the flip side is that CoDel can be implemented anywhere to be effective; the more places it is implemented the more effective it is. My understanding is that we can see some benefit without upgrading the entire internet.

By: Dave Täht

Dave Täht — Fri, 11 May 2012 00:33:08 +0000

Better explanations of how the algorithm actually works are being discussed on the codel email list. A start at one is here:

https://lists.bufferbloat.net/pipermail/codel/2012-May/000172.html

By: davidcollierbrown

davidcollierbrown — Thu, 10 May 2012 17:31:00 +0000

I’ll have to think about how to express what I don’t see… (not as hard as proving a negative, but sorta brain-warping (:-))

By: gettys

gettys — Wed, 09 May 2012 21:19:33 +0000

Unless you can explain what was unclear about them, it’s hard for Kathie and Van to clarify the diagrams. Remember, the CACM version is later this summer; it’s still possible for them to do such clarifications.

By: davidcollierbrown

davidcollierbrown — Wed, 09 May 2012 18:04:24 +0000

At the expanse of being slightly off-topic, I found the diagrams in the article a bit unclear. Since I’ve been following the discussion, that was a mere nuisance. For others, it could be less good. –dave

By: Dave Täht

Dave Täht — Wed, 09 May 2012 02:43:37 +0000

I almost put up a paypal link this week. Still may. I’m rather tired of living on top ramen.

By: Maciej Sołtysiak

Maciej Sołtysiak — Tue, 08 May 2012 20:32:13 +0000

Great stuff, Jim, thanks for covering this!

ps. You forgot to put a PayPal donation link there around financial support

By: gettys

gettys — Tue, 08 May 2012 19:25:32 +0000

That’s a really good question.

In short, any queue that could ever be a bottleneck needs management. So ultimately it needs to be in our operating systems (even on ethernet!), our home routers, our broadband gear, and everywhere else, including internet routers (where CoDel would be able to replace the existing *RED or other AQM algorithms).

We know we’ve got really bad problems in our OS’s, our home routers, broadband gear and 3g. Beyond that, the data gets spottier and less definitive, though convincing there are still lots of problems. In short, we’re draining an Internet wide swamp.

By: Bill Stewart

Bill Stewart — Tue, 08 May 2012 19:17:04 +0000

Where does CoDel need to be implemented to be effective? Only in the endpoints, only in intermediate points like routers, or both? Does it matter how much of the round-trip path is using CoDel as opposed to other queue management techniques?

Upgrading endpoints can be relatively easy (run a software update in Linux/Mozilla/etc.), and updating some routers and firewalls can be easy if the vendor updates their code, but others are hard to fix (like home DSL routers and cable modems.)

By: I, Cringely » Blog Archive Beginning of the end for bufferbloat - I, Cringely - Cringely on technology

Tue, 08 May 2012 17:42:54 +0000

[...] by Kathleen Nichols and Van Jacobson. If you read the paper please also look at today’s blog post from bufferbloat pioneer Jim Gettys explaining what it’s all [...]