dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
5
share rss forum feed


ropeguru
Premium
join:2001-01-25
Mechanicsville, VA
reply to pflog

Re: [Business] [POLL] Do you have the new 15/3 or 27/7 business

Looks like they have really fubar'ed something in the config now. When I run a shaper probe it is intermittently failing on too much packet loss. Pinging from the inside out I see the loss and when pinging from the outside to the default gateway IP on the SMC, it also starts throwing timeouts, which tells me the SMC is having issues.

Update: Even when pinging the WAN DHCP IP address on the SMC, it also starts timing out along with the other IP's. So this is definitely a SMC issue all around not just a router processing issue within the SMC.


pflog
Bueller? Bueller?
Premium,MVM
join:2001-09-01
El Dorado Hills, CA
kudos:3

1 edit
Click for full size
Looks like unfortunately the Netgear device doesn't help with this 27/7 issue. Although the speeds are more sawtooth and less erratic now with the Netgear, there's clearly still an issue with the speeds after power boost ends.

However, the tech said something about pulling the boot file and config from my SMC and then pushing one or both to the Netgear, so perhaps he wasn't supposed to do that, not sure.

On the plus side, my SNR was sitting between 35 and 35.5 on the 4 channels on the SMC, and this Netgear's 8 channels are sitting at 37+, so I can't complain about that.

*edit* Adding shaperprobe results:

Estimating capacity:
Upstream: 10800 Kbps.
Downstream: 35398 Kbps.

Upstream: Burst size: 9795-9915 KB; Shaping rate: 7429 Kbps.

Downstream: Burst size: 20208-20442 KB; Shaping rate: 28560 Kbps.

The tech brought up a good point. Powerboost fluctuates based on the load in the CMTS right? Could it be as simple as my CMTS thinking "powerboost is still uncapped" but the boot file is capping it? If the CMTS is "leaving the flood gates open", but the modem is clamping it, couldn't it cause exactly this behavior (packet loss, which causes the TCP stack to back off, then ramp back up, etc etc)?


espaeth
Digital Plumber
Premium,MVM
join:2001-04-21
Minneapolis, MN
kudos:2

1 recommendation

said by pflog:

The tech brought up a good point. Powerboost fluctuates based on the load in the CMTS right? Could it be as simple as my CMTS thinking "powerboost is still uncapped" but the boot file is capping it? If the CMTS is "leaving the flood gates open", but the modem is clamping it, couldn't it cause exactly this behavior (packet loss, which causes the TCP stack to back off, then ramp back up, etc etc)?

Not really. This is a rate shaping vs rate limiting problem. TCP automatically rate limits itself in relation to RTT and packet loss.

Rate shaping is about exploiting that function of TCP and implementing a "leaky bucket" mechanism where you start discarding packets even before the rate threshold is crossed. This causes TCP to back off on how quickly it tries to scale the ACK window and when you zoom out a bit on the IO graph you get that relatively smooth line you're accustomed to seeing.

Rate limiting is generally implemented as a strict threshold limit. All packets under the "bucket" threshold are accepted, everything above the threshold is dropped. This creates problems for TCP because each time it scales the window it sees 0 drops while still under the threshold which causes it to continue to be aggressive with the scaling. Eventually it scales pass the threshold, loses more than a few % of a congestion window's worth of packets, and kicks in congestion avoidance by cutting the TCP window to 1/2 of what it was. That process results in the sawtooth graph you see above.

Given that you're seeing the same behavior with 2 different CPE devices, I would speculate they are probably implementing the downstream shaping on the CMTS (this is logically the best place to do it anyway). I suspect either the CMTS profile for the 27/7 offering is incorrect, or the configuration may have found a code bug that causes rate shaping to not operate correctly with these specific parameters used.


pflog
Bueller? Bueller?
Premium,MVM
join:2001-09-01
El Dorado Hills, CA
kudos:3
Yeah it seems pretty clear that the CMTS and 27/7 profiles are not playing nicely together in some areas, while they're ok in others.

I'm not even sure who needs to address this at this point? It seems like a localized issue that the CMTS engineers would need to fix/update, but how we get their attention is unknown to me.

ComcastDJ See Profile and jlivingood See Profile were/are both looking into this, but they were hoping the new Netgear modem would fix it. I'm not sure if they have the means or "authority" to push for CMTS updates in various places or if that all has to be driven locally.
--
"Women. Can't live with 'em, pass the beer nuts." -Norm


ropeguru
Premium
join:2001-01-25
Mechanicsville, VA
Off topic, but does the Netgear allow for native IPv6? If so, are you seeing it available at your location?


pflog
Bueller? Bueller?
Premium,MVM
join:2001-09-01
El Dorado Hills, CA
kudos:3
said by ropeguru:

Off topic, but does the Netgear allow for native IPv6? If so, are you seeing it available at your location?

It's listed as not supporting IPv6 on the comcast site here:

»mydeviceinfo.comcast.net/

Don't see any IPv6 stuff in the Gateway Summary -> Network tab. But I imagine it does support it, just not configured/setup yet?
--
"Women. Can't live with 'em, pass the beer nuts." -Norm