Study: Bandwidth hogs aren't responsible for peak network congestion

[nismotigerwvu]

http://boingboing.net/2011/11/30/study-bandwidth-hogs-arent.html

Was a study even needed? To be honest, backhaul is A LOT like Interstate Highways. It's not the people who drive the farthest that cause congestion, its the rush of people driving home from work (who ironically go home and turn on netflix and tie right into my comparison) that do. The heavy users do their "damage" on off hours and really have no bearing on the service quality of everyone else.

 

[drebo]

It's a series of tubes, man!

 

[nismotigerwvu]

Ha, if this was reddit I'd give you an upvote for that. But more on topic, it seems like there are only two solutions for ISPs feeling the crunch at peak hours, either add more capacity (the logical solution) or trim customers. It all really boils down to X amount of capacity can adequately serve Y number of customers.

 

[paperwastage]

cough bittorrent through the night cough

but anyways, other people are getting their stuff done during peak hours, which could throttle/limit the bandwidth that the data hogs use(so they get effective 1mbps instead of 3)?

 

[her209]

Using your Internet connection during peak time slows down everyone else.

DERP DERP

 

[VirtualLarry]

I have 30/25 service through Verizon, and over turkey-day holiday, Speedtest.net showed I was only getiing 7Mbit/sec down.

 

[Railgun]

Bandwidth isn't a user issue, it's a provider oversubscription issue. Everyone wants the fastest thing out there. They get it, but forget that if everyone else does too, problems will ensue. I get 50/5 here through virgin media. Works great during non peak. During those peak hours though, fractions of a single Mb are seen.

 

[cmetz]

Hmm... researchers I've never heard of, wide PR campaign advertising this "research", directly conflicts with what I've been told by people I know and trust at real ISPs.

 

[nismotigerwvu]

Hmm... researchers I've never heard of, wide PR campaign advertising this "research", directly conflicts with what I've been told by people I know and trust at real ISPs.

Large aggregate usage != primary source of peak time congestion. As mentioned earlier in the thread, it is more a matter of overselling on the case of the ISP. I can't really blame them on that part, as financially it makes a lot more sense to serve slightly more customers than you can handle without any congestion as most minor slowdowns go unnoticed. The issue is when the ISP grossly oversells or even worse, tries to institute capped/tiered models that do nothing to to alleviate the issue. While some people may argue the "you want bandwidth you should pay for it" point, the counterpoint of "if you want hard earned dollars, you had better earn it" cuts just the same. Fiber is expensive, and increasing capacity is a non-trivial matter however, usage on a node is rather predictable on the long term and if it appears congestion is likely to occur, the sound business practice would be to limit, if not cut off the addition of new subscribers until it becomes profitable to expand userbases/capacity.

 

[Railgun]

The thought of fiber being expensive isn't really the case anymore. It's now the optics behind it. 100G, where these companies need to get on, is hugely expensive now. While I don't presume to know each providers infrastructure between a customer and the core of their network, I don't think it's at a state that it should or can be, obviously.

Now it's a chicken and egg scenario. Do they build it or do we keep buying it up to eventually fund it? I don't even want to touch profit margins that slow the rollouts.

 

[drebo]

Fiber is hugely expensive to deploy. Much more expensive than the optics used at either end. A 10 mile fiber run will cost a quarter million dollars. 40g or even 100g optics for either end don't come close to the cost of that fiber. Because, remember, the aggregators and muxes at either end function for many fiber runs, so their cost is split among all runs they connect with. Talking just about the optics on either end, and you're a fraction of the cost of the fiber itself.

 

[drebo]

To the story, oversubscription is a fact of life. Everywhere. It's the only way to bring multi-megabit speeds to everyone in the last mile. It's also the reason why a DS3, which is "only" 45mbit, is still several thousand dollars per month, while your 50mbit FiOS is only $50/mo.

You want dedicated bandwidth to everywhere? Get peering agreements with everyone.

 

[nismotigerwvu]

To the story, oversubscription is a fact of life. Everywhere. It's the only way to bring multi-megabit speeds to everyone in the last mile. It's also the reason why a DS3, which is "only" 45mbit, is still several thousand dollars per month, while your 50mbit FiOS is only $50/mo.

You want dedicated bandwidth to everywhere? Get peering agreements with everyone.

Oh I agree with you for the most part here. With a DS3 (or any other dedicated line) you are effectively paying the premium for just that, a dedicated line with a constant, guaranteed bandwidth. Oversubscription is only a problem when the congestion at peak hours is really pronounced (for the average user, they likely wouldn't notice until it sank below 3mbit or some obscene level of latency). The difference here is that the smart ISPs will be the ones that use the higher profit margins from this situations to fund network upgrades rather than the ones like Comcast and the like that just line their pockets and point fingers at the users when their outdated/overwhelmed networks caves in on itself.