Posted on November 27, 2017 at 2:37 pm

So why are the big ISPs fighting against Title II now?

I’m going to reuse acronyms and “framework” described in my Jul 15th article – ISP, CP, NSP.

The claim by many on the side of 17-108 is that Title II (keeping the current 2015 final order in place) represents chafing regulation. From the side of the ISPs like Verizon and Comcast the regulation basically requires them to treat everything the same best effort. Regardless of if it’s “on net” (Xfinity Video for Comcast f/ex), competing or not, or third party/off net (basically all the rest of the internet – Netflix, Google, Amazon, Hulo) — I can see their argument against it because it strongly prevents them from prioritizing any service for any reason. Leaving everyone at best effort (among other things yes)

What does that mean for an access ISP? Well it means you have to scale your access network towards the “worst case” scenarios. Everyone tuning into Netflix at 6PM. Everyone turning on the football game via the NFL streaming services. This is expensive for them because if they don’t scale their network so that these worst case scenarios they can still provide good best effort service to Google, YouTube and Facebook at the same time, they get lots of consumer complaints.

This means much more significant investment in their network, especially at the access layers, and into their core as well.

If they can prioritize traffic to services that are the squeaky wheels, or, that offer them more money in some form then the capacity scaling can be far, far more conservative. The answer to “my Netflix sucks on Monday night football nights” becomes “Oh then you need to pay $5.99 for the Netflix tier”

Why the push now? This actually likely has a lot to do with the hardware they’re buying now… I’ll use Juniper (JNPR) gear as an example as I’m most familiar with it but it doesn’t matter whose catalog you open, Cisco has similar product ranges and features, Huwei, etc. Back to JNPR Starting with the three most recent line card generations for their biggest routers they’ve gotten features with every single card that allows them to classify up to 32k different services. And that’s in the base. The just prior generations the feature existed in specialized cards and was much more limited. They can even upgrade it to a million odd separate classifications. These are specifically the MPC7E (32k base, 512k upgradeable classifications), MPC8E and MPC9E (both 32k base, upgradeable to 1m — base 10 not base 2 for the geeks that care) line cards. These bad boys on the low end will churn through and sling 400Gbit/sec (MPC7E) and 1600Gbit/sec for the biggest, baddest MPC9Es. These big animals live in their network core. It’s REALLY easy to deploy these features here, you’re touching a few hundred routers instead of thousands. Doesn’t require your access network to change substantially either, it in fact leverages stuff that has existed in the access network for a couple decades or more depending on how you actually implement it. A few thousand dollars in IT time. A few tens or maybe a hundreds of thousand dollars to implement the billing and support requirements to add/remove these tiers.

So how much $$$ ca$h $$$ are we talking about here? Well the “low end” MPC7E card will set you back something in the neighborhood of $250k. For the big daddy MPC9E … well … as the sayign goes if you have to ask you certainly can’t afford it. But if you’re not willing to talk about $1.2M mark well, don’t bother.  Oh and that $1.5M, that’s just the card.  You’ve got to plug in interface cards to actually connect that bad boy to anything. Yeah, let that soak in.  Somewhere “upstream” of you you’re using a piece of gear that costs easily $1.5M – actually you’re crossing over quite a few of these on your way out to Netflix or Facebook…or here to Dotblag.  And the three different chassis that fits those MPC9E cards will hold from about 8-20 of those bad boys. Easily $15M on the low side. Multiply that by your number of POPs (Points Of Presence) – you’re into small nation territory for capital expenditure.  In fact since I work at the DC that the Dotblag server lives in (ignore Cloudflare you geeks) I can tell you that no less than about $2M of kit stands between the 1Gbit port on my server and the connection to the first Comcast owned piece of gear.  So think about that.  Just putting this up there for you guys to see I’m leveraging *easily* around $10M in kit.

So now… say you’re in the Comcast board room. We’re talking financials.  You’ve just been handed the 10-K filing and you see this summary of how much you’ve spent on CapEx.  There’s probably a discussion there about exactly what that $6.5B line item entails…. Comcast CapEx on network gear is hard to pin down b/c they don’t really break it out…but given the total pie for their 2017 10-K filing is $6.5B … lets call it a cool $1B (which is not out of line here, it’s quite possibly far more than that) Then you and your buddies start talking about this number, and the CTO tells you that for another $250k, hell lets be generous and say it costs $15M And another $100M in lobbying to get the law back on your side all in $115M-$250M… you can net additional revenue of – for argument lets say your CFO figures it’s $5/subscriber/month – around 23m IP/internet subscribers — $115M/MONTH – $1.3B/yr, easily, year on year, by just using features that are now already there.  So we’re talking about an addition $1.1B current year, and $1.3B after…And that’s ignoring the ability to defer certain CapEx since you’ll now have to scale that core quite a bit less. Except you can’t. Because the law says you can’t.  Say you are Brian Roberts.  You’re the Comcast CEO.  You personally hold around 22m shares of Comcast stock.  Your share in that $1+B/yr additional revenue via dividends payouts could easily be $2m… What’ll you do?