[conspire] Comcast Business (ISP/Internet) and "their" equipment (3 and a half "nines" of availability?)

[Michael Paoli]

Well, got Comcast Business (ISP/Internet) installed 2019-12-12.

start and end of errors:
GW PING: ping 8.8.8.8 FAILED 2019-12-19T18:35:31+00:00
BALUG PING: ping -6 2001:470:1f05:19e::2 FAILED 2019-12-19T18:36:29+00:00
GW PING: ping 8.8.8.8 FAILED 2019-12-19T18:37:05+00:00
BALUG PING: ping -6 2001:470:1f05:19e::2 FAILED 2019-12-19T18:38:02+00:00
GW PING: ping 8.8.8.8 FAILED 2019-12-19T18:38:31+00:00
BALUG PING: ping -6 2001:470:1f05:19e::2 FAILED 2019-12-19T18:39:29+00:00
GW PING: ping 96.86.170.230 FAILED 2019-12-19T18:39:51+00:00
GW PING: ping 8.8.8.8 FAILED 2019-12-19T18:40:06+00:00
BALUG PING: ping -6 2001:470:1f05:19e::2 FAILED 2019-12-19T18:41:04+00:00
GW PING: ping 96.86.170.230 FAILED 2019-12-19T18:41:29+00:00
GW PING: ping 8.8.8.8 FAILED 2019-12-19T18:41:52+00:00
BALUG PING: ping -6 2001:470:1f05:19e::2 FAILED 2019-12-19T18:42:50+00:00

At present I'm using "their" equipment (fairly likely I'll change
that in future, but for the shorter term it was the quicker easier path
to get stuff up and running, and avoid any potential "finger pointing" as
to where any issues might be).

When the technician did the install, I asked him when rebooting the
cable "modem" ("their" equipment), about how long it took to get fully
back up and online.  He said (up to) about 5 minutes.
At the very first power-up and check out, initially all seemed fine, ...
then things stopped working for several (approximately 5?) minutes ...
he soon realized what it was doing - had downloaded fresh firmware, updated
and was rebooting (resumed functioning after that) - that was 2019-12-12.
And, as shown again, I see ... looks like (most) connectivity was
lost for a bit over 5 minutes.  Also, my Raw Bandwidth ADSL service
hasn't dropped yet ... but that's coming *real soon* (any time now).
So, yes, my program also still monitors my Raw Bandwidth connectivity,
most notably the associated gateway (GW) - didn't lose connectivity to
that, and that's at the remote end of the ADSL link.  As far as I can
tell - and guestimated by very low latency ping times, the Comcast Business
GW is, I believe, on the cable "modem" device.
The ping check of 8.8.8.8 in my wee 'lil monitoring program - at present
that just checks via default route (and default source IP) ... which at
present is via Comcast Business.  If I explicitly set source to
that of Raw Bandwidth (and likewise any return traffic from such), then
that (thus far still) goes via Raw Bandwidth.

So, in less than 10 days, already seem Raw Bandwidth has much higher
uptime/reliability, than Comcast Business.
If these "modem" reboots happen, uh, "automagically", and, oh, gee, maybe
about weekly (guestimating frequency from behavior thus far), that means
approximately 5 minutes of outage per week with Comcast Business ... on
"their" (bloated, overfeatured, and for better and/or worse very highly
capable cable "modem" device) device.  In over 16 years of service
with Raw Bandwidth, outages were exceedingly rare, and generally very short.

"Just sayin'".

Anyway, if/presuming I replace the Comcast Business provided hardware with
my own - not only will that save approximately $15.00/mo. (about $20.00/mo.
starting next year), with a much more minimalistic simplistic device,
and me controlling if/when any firmware updates are done, that would probably
significantly increase net ISP uptime performance over Comcast Business.
Oh, and ROI would be rather short too.  On 2019-12-12, I did quick
peek/search on-line.  Between used, and brand spankin' new, street prices
of *same* hardware (make & model) of what Comcast Business provides, as I
(approximately) recall, seemed to range from about $25.00 to about
$150.00.  So, not only short ROI, but even keepin' around a purchased
in-warranty spare for redundancy, would be cheaper than Comcast Business's
rental, and cost could probably be further reduced with a much less
feature bloated but sufficiently (minimalistic) capable device.
"Of course", with "their" device, they (also) support it, with
(my) "own equipment", their support ends at the end of the
cable (but I'd be fine with that).  And, too, could always
overlap for a short bit, to check and switch back and forth,
and make sure all the relevant setup/configure bits for "own equipment"
were fully determined and vetted and tested.
Anyway, somethin' for me to investigate further 2020Q1 (already more than
enough other stuff to keep me busy before then).

About 5 minutes per week (guestimated frequency) outages ...
$ echo '100-(5/(60*24*7)*100)' | bc -l
99.95039682539682539700
$
About 3 and a half "nines" of reliability/availability.
(Hey, if I really wanted to bother, could actually gather and count
up and measure the stats ... but not exactly all that high availability
of site and virtual machine (BALUG(/SF-LUG/BerkeleyLUG)) ... there are
also many other single points of failure (even if most of 'em don't
fail all that regularly).