Background: I am a network engineer by training and an engineering manager by trade who (on the side) owns a small business several states away that I manage remotely.
We are a business class customer on the highest tier (100/25) with a (self-provided) Cisco DPC3010 DOCSIS 3.0 8x4 Cable Modem. We had about 6 months of mostly-solid performance, but then back in January we started seeing serious (3-4%) packetloss on our first hop (from my firewall to the next layer 3 hop upstream). I run the business remotely, so anything more than 1% packetloss makes operating my business very difficult (VoIP, IP Cameras, POS, etc).
Packetloss over a 24h period looks something like this:
We experience recurring packetloss that usually lasts about 12 hours, then it will disappear for a while, and then it will show back up again around the same time of day. If you zoom out and look at a 60 day period, you will notice that the actual time-of-day when the packetloss occurs is also a rolling period (meaning it changes slightly from day-to-day):
It's kindof hard to see at this resolution, but basically over the course of about 4 weeks the periods of intense packetloss shifts from AM to PM.
I have called comcast support multiple times on this issue. Of course, each call starts with the rep first checking out the modem and verifying that they also see the packet loss, and they will verify that the levels on the modem are acceptable (they are). Then, they always try to tell me I need to rent a modem from comcast (untrue). Once we make it past that, they will claim it's a local wiring issue and dispatch a tech.
The tech will come on-site and verify that the issue is NOT our modem, our internal wiring, or the last mile from the pole to the demarc in the building. The tech will then tell us that it's an issue with the comcast network upstream, and the engineering department will need to look at it. During the last (third) visit, I gave the tech printouts of the packetloss graphs over 30, 60, 90, and 365-day periods. The tech agreed that the problem is obvious, and that he would provide the graphs to engineering to demonstrate the issue.
That's the end. Comcast never does anything about it -- no follow up, no additional contacts, no improvement in network performance. They literally seem to give zero cares about the situation, presumably because they understand I have no other options if I need > 20mbit/s uplink.
I can't really figure out a plausible root cause for this condition: if it's an oversubscription issue, the periods of packetloss would be mostly consistent (they roll over time); if it's a layer 1 issue (wiring), the periods would be near-constant (they're not -- we have long periods of stability) or tied to other weather phenomenon (no correlation of weather to performance).
So, I guess my questions to this forum are:
Thanks in advance...
I apologize for my delayed response and for the lack of interaction within the community. In my past experience when customers have connection issues while owning their equipment, pushing a temporary IPv4 only mode helps stabilize the connection. By pushing this mode it forces re-registration to the network usually correcting power and signal to noise issues. This process can take up to 40 minutes to complete if you would like to try this as a resolution. If you would like to take this option please let me know what time of day works best for you. (AM|PM)
Thank you for the reply! I would be glad to try this out as a temporary or permanent solution (at least for as long as I can get away with IPv4 only. ;-) PM works best for me, as we are open during the day.
Please let me know what to do to proceed.