- Tuesday, November 5, 2013

aqm@jabber.ietf.org

Tuesday, November 5, 2013< ^ >

Room Configuration

Room Occupants

GMT+0
[19:24:28] Dave Taht joins the room
[19:24:36] Dave Taht leaves the room
[20:01:38] g.white joins the room
[20:06:33] <g.white> test
[20:52:49] Guillaume Leclanche joins the room
[20:55:25] Dave Taht joins the room
[20:55:33] <Dave Taht> Hi greg!
[20:55:52] <Dave Taht> yes, we mentioned that the ns2 and ns3 models needed volunteer love. Were not aware of the status til your email.
[20:56:59] jlcJohn joins the room
[20:59:12] <g.white> great, yes I took a look at your slides, hence the status update...
[20:59:14] wesley.m.eddy joins the room
[21:05:22] ijjarvin joins the room
[21:21:14] <wesley.m.eddy> stuart cheshire is asking the question
[21:22:06] internetplumber joins the room
[21:22:12] <Dave Taht> Hi stephen!
[21:23:05] <g.white> D3.1 includes entirely new physical layer
[21:23:27] <g.white> early 2015 for initial products
[21:24:00] Lee Howard joins the room
[21:24:04] FJB joins the room
[21:24:25] Jason Weil joins the room
[21:25:04] Martin Stiemerling joins the room
[21:25:07] <g.white> also expectation is that the control path (drop probability update) is done in software, so can be updated easily in future
[21:25:43] <Dave Taht> yes, but I have no idea how to tweak that to make any sense in the future.
[21:26:40] <Dave Taht> I haven't benched the current pie vs fq_codel but in terms of if statements they are pretty close.
[21:27:14] <Dave Taht> Now that this stuff is stable I look forward to finally being able to evaluate it fully, try out various fq + pie variants, etc
[21:28:52] <g.white> the published ns2 modules will help in this regard
[21:30:25] <Dave Taht> http://www.cablelabs.com/specifications/CM-SP-MULPIv3.1-101-131029.pdf
[21:31:27] <Lee Howard> I missed something; what were the groans about?
[21:31:52] <Jason Weil> lack of ECN support
[21:31:54] <internetplumber> No mandate in the DOCSIS spec for ECN (I think)?
[21:32:36] <Lee Howard> ah. yeah, we need to do ECN. Not sure we need it required DOCSIS, per se.
[21:32:50] <Martin Stiemerling> right, it is not required in docsis
[21:32:59] <g.white> would love to see “early congestion notification”
[21:33:09] <ijjarvin> Has there been interest for "additional AQMs" already or is it just an option left open for future?
[21:34:50] <Martin Stiemerling> @ ijjarvin: post this question to the aqm list.
[21:35:10] <Dave Taht> well, certainly ARRIS presented some interesting stuff on the CMTS side in SCTE
[21:36:10] <Dave Taht> pie is mandated in cable modems. support for alternate aqms is baked into the spec, so those that want to innovate, with hardware powerful enough to innovate with, have options.
[21:38:53] <FJB> Early congestion notification is pretty doable with ECN marks; you can set the queue depth threshold very low knowing that it will not have a negative effect on the throughput of a session (which a loss has). Absent ECN, the signal (loss) has two effects - it stalls data flow as Stuart C described, and it also tells the sender to reduce its window. Greg, help us get ECN in there, and we can give you what you're asking for.
[21:39:21] <Dave Taht> ECN is supported in both pie and fq_codel on linux. Go play. Please.
[21:39:21] PasiS joins the room
[21:39:43] <FJB> That's not the question. The question is whether it's in the docsis spec
[21:40:01] Dan Wing joins the room
[21:42:25] <Dave Taht> Its not in the docsis spec because insufficient operational experience with it exists. Go get some, please.
[21:48:46] <g.white> ECN is hardware impacting for us. We’d need to have a *lot* of confidence that it gives benefit, does no harm, and isn’t going to change in the near future (we currently don’t have confidence in any of those).
[21:50:54] Martin Stiemerling leaves the room: Replaced by new connection
[21:50:55] martin.stiemerling joins the room
[21:51:21] <Dave Taht> hi martin
[21:52:13] <martin.stiemerling> Hi Dave
[21:53:21] <Dave Taht> Martin: have we ever actually managed to meet?
[21:53:32] <martin.stiemerling> nope, but we can do right after this?!
[21:55:12] <Dave Taht> excellent!
[21:55:20] <Dave Taht> I'm the guy with the eyepatch.
[21:56:52] cabo joins the room
[21:58:05] Jason Weil leaves the room
[22:00:31] martin.stiemerling leaves the room
[22:00:34] Lee Howard leaves the room
[22:00:50] Dave Taht leaves the room
[22:01:11] internetplumber leaves the room
[22:01:11] wesley.m.eddy leaves the room
[22:01:59] jlcJohn leaves the room
[22:02:11] cabo leaves the room
[22:05:22] FJB leaves the room
[22:05:39] g.white leaves the room
[22:05:41] Dan Wing leaves the room
[22:07:04] FJB joins the room
[22:10:11] PasiS leaves the room
[22:15:46] FJB leaves the room
[22:16:20] internetplumber joins the room
[22:17:45] internetplumber leaves the room
[22:18:22] cabo joins the room
[22:22:37] Dan Wing joins the room
[22:22:38] martin.stiemerling joins the room
[22:22:40] martin.stiemerling leaves the room
[22:29:44] wesley.m.eddy joins the room
[22:31:41] wesley.m.eddy leaves the room
[22:34:40] cabo leaves the room
[22:40:50] Lee Howard joins the room
[22:43:35] andrebeliveau joins the room
[22:50:32] Dan Wing leaves the room
[22:55:09] FJB joins the room
[23:00:24] andrebeliveau leaves the room
[23:17:31] PasiS joins the room
[23:20:49] PasiS leaves the room
[23:36:20] Lee Howard leaves the room
[23:40:34] FJB leaves the room
[23:42:37] Lee Howard joins the room
[23:52:16] Lee Howard leaves the room