vCMTS unloaded latency and jitter

plat2on1 · 2025-06-08T00:29:09+00:00

Finally got the “next gen” upgrades here after 2 days of downtime, and it’s absolutely terrible. We get years of LLD “trials” but they just roll out this garbage like it’s nothing? Left side unloaded, Right side upload loaded with 30 Mbps with iperf.

MacLeech

plat2on1 CM8200 doesn’t support LLD, Comcast only supports LLD on a select few “next gen” modems.

I vaguely remember other modems being supported during earlier trials, but can only dig up the S33 and CM1000v2 besides the XB7 and XB8 currently.
https://corporate.comcast.com/stories/comcast-kicks-off-industrys-first-low-latency-docsis-field-trials

plat2on1

MacLeech

This is the latest info for COAM modems and LLD.

MacLeech

plat2on1 So the S34 should have support by now if you’re on a vCMTS.

Do you have an OFDMA upstream channel active on the S34?

plat2on1

MacLeech Yep, getting the full 350+ megabit upload.

cypherstream

plat2on1 was this 8 PM last night you put this in? Maybe it needs a firmware update. Maybe it needs to soak for 48+ hours to find the update. What about power cycling it?

plat2on1

cypherstream firmware updated to AT01.01.010.042324_S3.04.735 when it was first turned on.

07/30/2025 18:20:34 Notice SW download Successful - Via Config file
07/30/2025 18:19:58 Notice SW Download INIT - Via Config file

MacLeech

plat2on1 Do you still see the difference in unloaded vs loaded latency with the S34?

satyr

hex0101

It appears that efforts are being made to reduce idle latency. The big question is, when will it be deployed?

plat2on1

MacLeech Yes but it doesn’t work as well.

cypherstream

satyr Oh gosh the op on that post, spot on. Upvoting, and its why I never switched from SECV. Also at work I’ve used many ISP’s and Comcast routes all over the state.
SECV DOCSIS 12ms
SECV FTTH 3ms
COMCAST DOCSIS 34ms
Windstream Kinetic 6ms
Crown Castle (lighttower) fiber 6ms
Starlink 40ms
Verizon Fios 8ms
Cogent 4ms

hex0101

cypherstream

Jason trying to offer assistance, and OP was being a fool. Either way, I understand their issues.

I have been gaming on my Comcast connection for over a decade and have never had any issues with the network. 40-50 ms avg ping in game. Could it be better? Yes. Is it bad? Not at all. Like everything else, it is going to depend on where you are geographically and where the server you are connected to is located and the peering arrangements between the ISP’s and backbone providers, YMMV. I have seen some people get 2.5 ms ping to Google on Comcast Metro-E, while others are getting 11 ms (20-21 ms on DOCSIS).

AT&T Fiber does shave off of a few ms, noticied that when running simple ping tests from my cloud server located 150 miles away from me (testing idle latency, and only 2 ASN’s total on the traceroute for the Comcast IP, and 3 for AT&T, yet AT&T has lower latency because of no DOCSIS last-mile). I don’t expect that to change until PGS is a thing, and even then, it may depend on how it is configured. PGS also takes away upstream bandwidth.

videomatic3

hex0101 PGS also takes away upstream bandwidth.

i have 2 gbps symmetrical. i am 1000% positive i wont be missing out on the 3mbps upload pgs will be using at its maximum before diminishing returns.

hex0101

videomatic3 I thought it was per modem on the node, hence taking away from the overall node bandwidth capacity. If it is not, then I 100% agree with you. I won’t miss 3 Mbps either even on a 300 Mbps plan.

videomatic3

hex0101
im not sure, but even if it was 3mbps per modem, on a node like mine theres a max of 80 houses, (n+0) who knows how many of those have comcast, 80×3=240. i could see how it adds up quickly. but even then theres about 20 houses per leg and it can get divided by that too.

fmd

Every grant that PGS issues is not counted as wasted bandwidth - only the grants that go unused, because the timing is wrong (the modem doesn’t have anything to transmit at the time that grant x is due). The CMTS is allowed to optimize, if no LL traffic is coming from the modem, to save bandwidth. The wasted bandwidth is difficult to predict.

We don’t yet know how Comcast will enable this feature. If it becomes upon request only, the amount of PGS modems will probably be cut way down.

videomatic3

fmd lets say it uses 3mbps of wasted to keep ahead of the modem, it will always be 3mbps wasted to keep ahead of the modem demand, it wont matter how much traffic it is or isnt using. if the modem demands 0mbps 3mbps will be wasted, if the modem is demanding 2000mbps, 2003mbps is granted. of course this will happen at a much more micro level and change every microsecond, but even without PGS changing the grant system from 2ms to less than 1ms will make an astronomical difference.

fmd

videomatic3 Look at it like this: If the modem has data to transmit, it is entitled to some of the shared bandwidth. A regular grant or a proactive grant, it doesn’t matter.

Let’s say the CMTS sends a grant to the modem each ms. The timing is perfect, so the modem uses every grant and all of it. Nothing is wasted. Without the proactive grants, the modem would still get the grants, just with a different timing.

If that is the PGS config, it is required to send grants at that rate and size (approx). It is free to send more, if there is available bandwidth.

lesmikesell

fmd Look at it like this: If the modem has data to transmit, it is entitled to some of the shared bandwidth. A regular grant or a proactive grant, it doesn’t matter.

It makes sense for ‘your’ modem to be able to fill the proactive grant with the highest priority thing it has to fill it, but it will still be a timeslot stolen from all the other modems on the some line - and your modem will lose all the proactive grant time they get.

fmd

lesmikesell Prioritized grants may not be the correct term. The LL traffic has the same DOCSIS priority as the classic traffic. The proactive grant only gives you a latency advantage.

Let’s say your speed tier is 300 Mbps and you fill your LL queue with that amount of traffic. Is PGS going to preallocate that to you? No (the CMTS is free to do it, if nobody else is using the bandwidth). If the proactive grants are not sufficient, the modem must use the request method, like always.

The issue is congestion events. Many wasted grants are bad in that context, because they cannot be re-allocated to other modems. The amount of upstream bandwidth, PGS config and number of modems take part in this equation.

lesmikesell

Won’t that give somewhat misleading ping test results? That is if you test with a mostly unloaded line you’d see time results from the immediate send but when you are actually moving data you’ll mostly have to wait the roundtrip for a grant.

« Previous Page Next Page »