Dane DeValcourt

As someone who has spent 15 years, entire career now, as a network engineer for both enterprises and for ISP’s it’s difficult to have a single opinion on some things when it comes to networks.   

I really do see and understand both sides of most arguments around net neutrality. When I take off my network engineer hat and put on my consumer hat I wouldn’t be too happy with a lot of this stuff either.

However, when I sit back and think about things rationally its just not so black and white to me.

I know that often when analyzing network traffic the bulk of that traffic will be due to a few offenders.

This sort of thing reminds me of the days in grade school where you had a few bad students creating havoc and thus affecting the whole class in some way.  It always seemed unfair why the class would get punished for the actions of a few.

People want to do voice, video and data over the network these days but they don’t want the ISP’s to interfere with their traffic.  On private LANs and WANs we do the same thing (voice, video and data) but we also know that in order to do this successfully we have to employ quality of service (QoS).  Without QoS voice quality would be very unpredictable and video would be choppy.

The way QoS works is by looking at all the traffic coming in, orgainizing the traffic by various priorities and then allowing the highest priority traffic to leave first.  What often happens is the lowest priority traffic might get dropped.  This is typically ok because most traffic can tolerate being dropped as the sending device will get notified that some of its data was dropped and it needs to resend it.

Applications like voice and video however do not tolerate dropped traffic very well.  If you drop voice traffic the user will notice this change because they are actively listening.  People don’t notice dropped packets when surfing the internet because the retransmits of this dropped data happens so quickly and at worst it may just mean the website they were going to loads up a little slowly.

So you have these people out there who believe that ISP’s should simply provide a dumb pipe with no QoS or anything since that would interefere with or manipulate their traffic in some way.

The problem with dumb pipes is that ISP’s rely on the principle of ‘over subscription’ in order to make money.  You simply cannot be profitable as an ISP without oversubscribing.

Consider this, here are some of the largest pipes ISP’s can get to the internet.  

OC-48 = 2.5G = 2488Mbit/s  divide by 10Mbit/s = 248 users

OC-192 = 10G = 9953.28Mbit/s  divide by 10Mbit/s = 995 users

The reason I got on this whole topic was because of some things I heard at this weekends CampFiber meeting.  Those living in Lafayette, Louisiana know that the local utilities company (LUS) is building out a fiber to the home solution for the cities residents.  LUS has been in the fiber business for some time providing service to businesses.  According to the info on their website (http://www.lusnet.net/) they have two DS3′s to the internet.  DS3′s are roughly 45Mbit/s, so combined were looking at about 90Mbit/s of capacity to the internet.

Again I have to say I do not know how old this data is and I do not know if these same circuits to the internet will be leveraged by their new home user service.  This would be good info to know though and I would love for anyone who does know to inform me of the truth.

OK, so we have established LUS has 90Mbit/s of internet connectivity.  This is currently being used by their business customers and by the City of Lafayette as well.  And it could potentially be used by their home customers but that is not a fact yet.

LUS Fiber is advertising that their minimum internet package will be 10Mbit/s for home users.   It’s also important to note that based on this document that they estimate their subscriber count will be 28,500.  The document states a pass count of 57,000 potential customers and sub count of 28,500.  For those that don’t know what this means its pretty simple, pass count is who could potentially be a customer because the service passes close enough by them.  Sub count is the actual subscribers.

Lets give LUS the benefit of the doubt here.  Lets make the sub count lower than their expectations, so lets round down to 20,000.  Each customer gets a minimum of 10Mbit/s.  So if every user actually used the full 10Mbit/s we would be looking at 200,000Mbit/s.   And they have 90Mbit/s of internet bandwidth?  Thats a bit oversubscribed.

Now anyone in the business knows thats not going to happen.  You will never see all users fully utilizing their internet connection.  If 90Mbit/s is their total bandwidth and each user gets 10Mbit/s it would only take 9 users out of 20,000 fully utilizing the 10Mbit/s to max it out.  That is very possible however.

Now lets really give LUS the benefit of the doubt there.  Lets say they have an OC-192 connection to the internet.  So, thats roughly 10,000Mbit/s (its actually more like 9,000 but just making the match easy and giving them the benefit of the doubt here remember).   So we take 10,000Mbit/s and divide by 10Mbit/s and we get roughly 1000 users who could be running a full 10Mbit/s.  Thats not too bad there.  Thats like a 20:1 subscription ratio.  BTW, I SUCK AT MATH so please correct the basic math if I am wrong but realize I am generalizing here and not being exact.

So far I have only discussed home subscribers using the LUS internet connectivity despite the fact that businesses use it as well.  If business customers and home customers will share the same internet backbone that LUS has then my guess is that LUS will at a minimum seperate traffic into two priorities, one for business users and one for home users.  And it would make sense to give the business users higher priority then the home users so that home users can’t impact the business users internet connectivity.

If all of what I said is true (highly unlikely) then would this be deemed a non net neutral situation for the home subscribers?

I ask because Chance with LUS stated at CampFiber they had to be careful to not get into a net neutrality situation so they were just giving dumb pipes and thats it.  And if this is true, based on the numbers above which again were very generous it wouldn’t take much for a few subscribers to effect the internet performance of all subscribers without some means of QoS.

Which brings me back to the issue of whether network management by using any sort of QoS create a net neutrality issue?

Lets say you have a 1Mbit/s pipe to the internet.  You have two users who share this pipe but you sold them each 1Mbit/s.  For the sake of argument lets say that User A is doing streaming video that takes up the full 1Mbit/s pipe.  User B starts to do something else but only requires 500Kbit/s from the pipe.  Obviously something has to give here.  If you QoS things so that during peak times no one can get more than 500Kbit/s so that one person does not impact the other well then your not living up to what you sold them.  If you give video a higher priority then you aren’t being very neutral since you will impact what user B is doing effectively saying that user A’s traffic is more important then user B’s traffic.

How can one be “net neutral” in an over subscription business model?  You either can’t over subscribe, or hope that your aggregate traffic never exceeds your biggest pipe.

Looking for enlightenment here, would love to be shown the err in my thinking about this.