On Mon, Nov 26, 2018 at 1:56 PM Michael Welzl <michawe@ifi.uio.no> wrote:
Hi folks,
That “Michael” dude was me :)
About the stuff below, a few comments. First, an impressive effort to dig all of this up - I also thought that this was an interesting conversation to have!
However, I would like to point out that thesis defense conversations are meant to be provocative, by design - when I said that CoDel doesn’t usually help and long queues would be the right thing for all applications, I certainly didn’t REALLY REALLY mean that. The idea was just to be thought provoking - and indeed I found this interesting: e.g., if you think about a short HTTP/1 connection, a large buffer just gives it a greater chance to get all packets across, and the perceived latency from the reduced round-trips after not dropping anything may in fact be less than with a smaller (or CoDel’ed) buffer.
I really did want Toke to have a hard time. Thanks for putting hisback against the wall!And I'd rather this be a discussion of toke's views... I do tend tothink he thinks FQ solves more than it does.... and I wish we had asound analysis as to why 1024 queuesworks so much better for us than 64 or less on the workloads we have.I tend to think in part it's because that acts as a 1000x1rate-shifter - but should it scale up? Or down? Is what we did withcake (1024 setassociative) useful? or excessive? I'm regularly seeing64,000 queues on 10Gig and up hardware due to 64 hardware queues andfq_codel on each, on that sort of gear. I think that's too much andrenders the aqm ineffective, but lack data...but, to rant a bit...While I tend to believe FQ solves 97% of the problem, AQM 2.9% and ECN .09%.
I think the sparse flow optimization bit plays a major role in FQ_CoDel.
BUT: Amdahls law says once you reduce one part of the problem to 0,
everything else takes 100%. :)
it often seems like me, being the sole and very lonely FQ advocate
here in 2011, have reversed the situation (in this group!), and I'm
oft the AQM advocate *here* now.
Well I’m with you, I do agree that an AQM is useful! It’s just that there are not SO many cases where a single flow builds a standing queue only for itself and this really matters for that particular application.
But these cases absolutely do exist! (and several examples were mentioned - also the VPN case etc.)
It's sort of like all the people quoting the e2e argument still, back
at me when dave reed (at least, and perhaps the other co-authors now)
have bought into this level of network interference between the
endpoints, and had no religion - or the red in a different light paper
being rejected because it attempted to overturn other religion - and
I'll be damned if I'll let fq_codel, sch_fq, pie, l4s, scream, nada,
I admit to getting kind of crusty and set in my ways, but so long as
people put code in front of me along with the paper, I still think,
when the facts change, so do my opinions.
Pacing is *really impressive* and I'd like to see that enter
everything, not just in packet processing - I've been thinking hard
about the impact of cpu bursts (like resizing a hash table), and other
forms of work that we currently do on computers that have a
"dragster-like" peak performance, and a great average, but horrible
pathologies - and I think the world would be better off if we built
more
+1
Anyway...Once you have FQ and a sound outer limit on buffer size (100ms),depredations like comcast's 680ms buffers no longer matter. There'sstill plenty of room to innovate. BBR works brilliantly vs fq_codel(and you can even turn ECN on which it doesn't respect and still get agreat result). LoLa would probably work well also 'cept that the gittree was busted when I last tried it and it hasn't been tested much inthe 1Mbit-1Gbit range.
But corner cases aside, in fact I very much agree with the answers to my question Pete gives below, and also with the points others have made in answering this thread. Jonathan Morton even mentioned ECN - after Dave’s recent over-reaction to ECN I made a point of not bringing up ECN *yet* again
Not going to go into it (much) today! We ended up starting anotherproject on ECN that that operates under my core ground rule - "show methe code" - and life over there and on that mailing list has beenpleasantly quiet. https://www.bufferbloat.net/projects/ecn-sane/wiki/.I did get back on the tsvwg mailing list recently because of someludicrously inaccurate misstatements about fq_codel. I also made astrong appeal to the l4s people, to, in general, "stop thanking me" intheir documents. To me that reads as an endorsement, where all I didwas participate in the process until I gave up and hit my "show me thecode" moment - which was about 5 years ago and hasn't moved on theneedle since except in mutating standards documents.The other document I didn't like was an arbitary attempt to just setthe ecn backoff figure to .8 when the sanest thing, given thedeployment, and pacing... was to aim for a right number - anyway.....in that case I just wanted off the "thank you" list.
So let’s draw a line between L4S and “the other document you didn’t like”, which was our ABE.
L4S is a more drastic attempt at getting things right. I haven’t been contributing to this much; I like it for what it’s trying to achieve, but I don’t have a strong opinion on it.
Myself, I thought that much smaller changes might have a better chance at getting the incentives right, to support ECN deployment - which was the change to 0.8.
Looking at our own document again, I am surprised to see that you are indeed in our acknowledgement list:
We added everyone who we thought made useful suggestions - it wasn’t meant as a sign of endorsement. But, before RFC publication, there is still an opportunity to remove your name.
=> I apologize and will remove you.
I like to think the more or less rfc3168 compliant deployment of ecn
is thus far going brilliantly, but lack data. Certainly would like a
hostile reviewers evaluation of cake's ecn method and for that matter,
pie's, honestly - from real traffic! There's an RFC- compliant version
of Pie being pushed into the kernel after it gets through some of
stephens nits.
And I'd really prefer all future discussions of "ecn benefits" to come
with code and data and be discussed over on the ecn-sane mailing list,
or *not discussed here* if no code is available.
You keep complaining about lack of code. At least for ABE:
- I think the code is in FreeBSD now
- There is a slightly older Linux patch. I agree it would be nice to continue with this code… I don’t have someone doing this right now.
Anyway, all code, along with measurement results, is available from:
, but… yes indeed, being able to use ECN to tell an application to back off instead of requiring to drop a packet is also one of the benefits.
One thus far mis-understood and under-analyzed aspect of our work isthe switch to head dropping.To me the switch to head dropping essentially killed the tail loss RTOproblem, eliminated most of the need for ecn.
I doubt that: TCP will need to retransmit that packet at the head, and that takes an RTT - all the packets after it will need to wait in the receiver buffer before the application gets them.
But I don’t have measurements to prove my point, so I’m just hand-waving...
Forward progress andprompt signalling always happens. That otherwise wonderful piecestuart cheshire did at apple elided the actual dropping mode versionof fq_codel, which as best as I recall was about 12? 15ms? long andtotally invisible to the application.(I think people easily miss the latency benefit of not dropping a packet, and thereby eliminating head-of-line blocking - packet drops require an extra RTT for retransmission, which can be quite a long time. This is about measuring latency at the right layer...)
see above. And yea, perversely, I agree with your last statement.
Perversely? Come on :)
A
slashdot web page download takes 78 separate flows and 2.2 seconds to
complete. Worst case completion
time - if you had *tail* loss would be about 80ms longer than that, on
a tiny fraction of loads. The rest of it is absorbed into those 2.2
seconds.
Yes - and these separate flows get their own buckets in FQ_CoDel. Which is great - just not much effect from CoDel there.
But I’m NOT arguing that per-flow AQM is a bad thing, absolutely not!
EVEN with http 2.0/ I would be extremely surprised to learn that manywebsites fit it all into one tcp transaction.There are very few other examples of TCP traffic requiring a lowlatency response. I happen to be very happy with the ecn support inmosh btw, not that anybody's ever looked at it since we did it.And I'd really prefer all future discussions of "ecn benefits" to comewith code and data and be discussed over on the ecn-sane mailing list,or not discussed here if no code is available.BTW, Anna Brunstrom was also very quick to also give me the HTTP/2.0 example in the break after the defense. Also, TCP will generally not work very well when queues get very long… the RTT estimate gets way off.
I like to think that the syn/ack and ssl negotation handshake underfq_codel gives a much more accurate estimate of actual RTT than weever had before.
Another good point - this is indeed useful!
I don’t know… in academia, the mix of really contributing to the kernel on the one side, and getting academic results on the other, is a rare thing.