P2P Bandwidth Hogging the Net

zymano writes "zdnet has this article about bandwidth hogging p2p." I'm sure we'll see more rate limiting in the future and per-gig charges. The article says 60% of ISPs bandwidth is P2P, and that seems high to me, but not unrealistic. Besides, since most broadband is pretty seriously hamstringed in the upstream department, I'm not sure where they can go with this.

Comments

It's a realistic number

[oneiros27]

As an employee at a university, I can tell you that in fact, those numbers are realistic.

Unfortunately, with the port-hopping ability of some of the newer p2p networks, restricting their usage, or giving them a lower class of service than other protocols is exceedingly difficult.

The real problem in our case is not so much the people downloading, but as we have a rather fat pipe to the internet, we're seen as very favorable download farm for people to grab files from.

This is good news for telecom

[Kombat]

As a former Nortel Networks employee, I am glad to hear this type of news. Part of the whole reason for the telecom meltdown was predicted demand that never materialized. The growth of traffic was unfolding as expected, but in a quest for better profits, the telecom companies decided to curb demand instead of increasing supply. So instead of expanding backbones, they capped downloads.

They can only do this for so long. With the rollout of large-scale gaming networks like Sony's and Microsoft's (for the X-Box), the demand will keep growing, one way or another. Sooner or later, the Qwests and MCIs are going to have to bite the bullet and buy some terabit optical switches. They're going to have to open up their wallets, and then we should start seeing a rebound in the high-tech market.

So support your high-tech buddies! Saturate your network connection, make your ISP feel the bandwidth pain, nag them to upgrade! :)

true

[Tom]

It's a fact. I work for an ISP. 60% is a conservative figure, we've seen more than that at times.

Thing is: P2P wastes tons of bandwidth. The continuous searches, all the broken or incomplete downloads, not even to speak of the overhead.

Re:true

[throwaway18]

>P2P wastes bandwidth. As in efficiency. How much bytes traverse
>the wire to download 100 MB via FTP? via HTTP? via scp?
>Compare that to P2P.

It vaires considerably been filesharing systems. My rough guess based on watching various p2p clients, including primary/supernode traffic, not including TCP overhead, assuming long term use is

gnutella/0.4 1-2Gbyte traffic for 100Mb download
winmx/kazaa/G2MP 150-250MB
edonkey 120MB

If you compare to the www you need to consider how many
web searchs and pages people have to go to to find the file or information they want.

Alternative per-GB charges.. but then there's eBay

[adzoox]

What about me, I work from home (mostly) listing on eBay and responding in forums such as these to promote my business through association and have a website that I frequesntly upload to. (Note: I do not use my broadband as a server) I also occasionally use P2P and I have to download software updates for my customers from Apple and VersionTracker often ...

I'd prefer not to hear the ... it's a business expense that you must pay arguement. I have built my business on the model I am in right now at the prices I pay right now. It works perfectly. If ISPs want more bandwidth then there's optimizations and measures for stopping scam site hosting and SPAM mail that they could do.

A post above had a good point - should TV's cost more because they use up a significant portion of the electricity - after all, more electricity, more damge to the environment and more cost to the Power company!

it's not redundant, it's spare capacity

[RMH101]

it's not free, either. if it all gets used up, what happened to the network's forward planning?

60% Usage

[mustangsal66]

We have a 45Mb DS3, we are a cable modem service provider. Watching the traffic I can confirm, that from about 3pm until 10pm 60+% of our traffic is from P2P clients. Thats only the traffic we can track. Kazaa 2 can use port 80, and only gets reported as web traffic.

I see kazaa 2 traffic mostly. but also edonkey, kazaa 1, napster, and others.

Less then 1% of our users use 85% of the bandwidth. They're alloted 1Mb/s download, and they use it constantly.

Seems high

[Kebinu]

I work for a midsized ISP in Michigan and we monitor information about our bandwith usage and distribution by service/protocol with netflow. We use this information for capacity planning, verifying abuse complaints, billing, etc.

Using netflow data from our core routers, we estimate peer to peer traffic to be 1/3 of our bandwidth utilization. This is based on the assumption that most p2p applications run on their default ports. I believe this to be a fairly safe assumption since we do not police our customers or restrict them from using p2p apps. The number one thing we could do to reduce our bandwidth usage right now would be to install transparent caching services for HTTP, some 50% of our bandwidth is HTTP traffic.

Economics of IP bandwidth cost

[puzzled]

I've posted on this so many times I've written it up and placed it in my journal

http://slashdot.org/~puzzled

Here it is again briefly:

A T1 has 24 x 64kb channels. Getting one from a top level provider like Sprint or UUNet will cost about $1000/mo or $40/channel/month.

A 256kb DSL link is four channels and costs about $40/mo. Four channels times $40/mo = $160/mo cost for the ISP. I realize the average math skills of slashdot readers are about eighth grade level, so I'll finish it for you - $40/mo revenue minus $160/mo cost = -$120/mo. This is what happens to ISPs when people doing file sharing of any sort leave their retail connections running 24x7 and consume bandwidth in a wholesale fashion.

Its not about the MPAA or RIAA, evil scumbags that they may be, its just simple cost that is going to do in file sharing. Stop being a whiny end user and pay for some quality bandwidth, or shut the *(&@$(%*&@#$% up about it already.

Re:Two words: Metered Bandwidth

[-brazil-]

Would you pay metered costs on a landline phone?

I am.

Probably 98% of the people pay a flat fee for local calls

Um, no. Not outside the US. In most countries, metered local calls are the absolute norm. Which makes flatrate broadband internet access all the more attractive.

Actual bandwidth usage of an ISP

[slug359]

Here's a bandwidth chart of a medium sized UK isp: Zen, P2P is a pretty sizable chunk of the whole with HTTP not far beind.

Re:Actual bandwidth usage of an ISP

[joeboo]

Ellacoya Networks also makes a switch that, like the packeteer device, uses signature detection to determine if a flow is p2p or not.

Before we started to use their switches, p2p traffic accounted for 65% of our network traffic (both inbound and outboud). By placing p2p limits on the upstream (or outbound) portion of internet pipes, we were able to reduce that to 40%.

When Ellacoya introduced signature detection in their 5.0 release, we were able to catch Kazaa2 flows which reduced our p2p usage to 30%.

Not only has this saved us a tremendous ammount of money (we don't have to keep buying pipe), but we were able to keep the cost down for our customers that don't use p2p, or care.

Another use for the switch is usage-based billing. People scoff at usage based billing, but it evens the billing field. Look at the grandmother that uses her connection to send emails to her kids, and then look at the kid down the street. Both pay the same ammount, but one uses 95% more bandwidth. A recent study that we did here concluded that 5% of our users use 90% of the available bandwidth.

So, what is fair? Limiting bandwidth-hungry applications to the point that they are no longer useable, or charging the customer that wants to use that application his fair share of the costs of delivering the service?

stupid complaints.

[SuperQ]

Back in the day....

I remember back in 94-95 era, people started to complain that "this web thing is eating all our bandwidth" People complained that it was slowing down email, usenet, and IRC. It was a hog, all those images.. yada yada yada...

Now that the average connection is a factor of 10 faster than it was in 95, someone invented an application to utilize that bandwidth.. *SHOCK*

One of our local campus admins talks about how we double our campus backbone connection every 2 years.. we had about 200mbit when i started 2 years ago, now the plan is to get a 3rd provider, and jump to 400mbit before the next school year starts. _IF_ we can afford it. (yay for state budjet problems)

In another 5 years, people will forget all about P2P, because it will be background noise compared to the ********* protocol. Whatever they think up when 100mbit fiber starts to get rolled out.

If you want to really see what is in the works, look at Internet 2 projects. Our campus has 655mbit to I2, and it's already too slow for some of the research. Plans are in the works for a few gbits.

Re:I'm cornfused

[Tazzy531]

The ISP business is quite similar to how airlines overbook flights assuming people will not use their allotted amounts. There will always be people that use more and a lot of times, the connection will lie dormant. If you were to pay for the actual charge of having all your bandwidth, you'd be paying a lot more than $50 a month.

The DSL provider here does traffic-shaping...

[Kjella]

In short, the "cheapo" alternative here is .. hmm I don't remember exact speed, think it's 1024k/256k but at a max of 1Gb/mo, if you pass that you are throttled to 64k (ISDN speed) or can buy more (at a premium). Personally I got the 1024/256k/unlimited use at $100/mo, and I don't feel any shame over running it at 100% 24/7 either.

They're very up front about the metering, and it *is* much better than pay-per-minute over phone. Their calculation of how many pages you can view per month can't possibly include any lame flash sites, but other than that it's a straight offer. However, the other big competing company basicly said something like "we'll never offer metered connections, unlimited all the way" and they've earned a lot of customers on that, not sure how many *profitable* customers though...

Kjella

Cornell Univeristy Stats (with graphs, etc)

[ron_ivi]

http://www.cit.cornell.edu/computer/students/bandw idth/charts.html

Cornell had 60% outgoing and 50% incoming traffic as filesharing. Lots of pretty graphs, etc. on those pages.

Re:I'm blocking p2p on my network

[stinky wizzleteats]

Nice troll. I do have one bullet left in my mod gun right now, but I have decided to let you off with this warning.

Not everyone who uses p2p is illegally sharing copyrighted works. I have a p2p node that stays slammed offering completely legal and non-porn content. I host linux distros, stuff related to Orbiter space flight simulator (free), and stuff that gets slashdotted. (People still download the Starship Exeter videos).

The node runs slammed 24/7, and I've had to implement traffic control to be able to concurrently use my connection for other things. Why do I go to the trouble? Because p2p is the best hope of ordinary people to share information. The next Thomas Paine probably will not have access to a web server, and if he did post the 21st century version of "Common Sense", it would get DMCA'ed or shut down by the government. The only hope, then, of free speech is going to be p2p, particularly next generation encrypted/anonymous p2p networks.

If I'm smart enough to figure out how to shape traffic so that I can ssh over a 200k upstream connection swamped with p2p traffic, then I'm reasonably certain ISP managers can probably figure it out without finding new and creative ways to tax the first amendment.

Lots of dark fiber

[charnov]

Two comments:

1.) Thanks to WorldCom inflating growth figures (that's what got them into trouble) for nearly 10 years, there is a tremendous amount of fiber lines just sitting there doing nothing. Don't believe the hype, there is enough base infrastructure in the US to give every body a T1 or better (but then we wouldn't need phones, cable/satellite TV, radios, etc...heh). Wireless meshes are popping up all over the place (in cities anyways) that also can allow joe average to distribute broadband content (within the mesh). The next 10 years, eveything will shift to some form of wireless (just wait til the RIAA and pals start going after spectrum rules...man the fur is gonna fly)

2.)If they (Broadband ISPs) want to control traffic, just sell service with a QoS agreement. I would rather have a business line (at the same price I have a consumer line) with 24x7 guarantied bandwidth at a lower rate than I have now for download (say 768/768).

Whoops, on the subject of spam. The last company I worked for spam cost the company over $2 million dollars a year in bandwidth (hard to filter BEFORE it hits your gateway).

Why an IP T1 still costs $1k/month

[puzzled]

Some whiny end user type responded to my post about the economics of being an internet provider with a brief rant on how technology should have brought the cost of a T1 down a long time ago.

Lets investigate the reality behind a typical Sprint T1 install at $1,000/month.

A T1 is composed of several components, the first being the local loop to the CO. You've got two or four copper wires buried in the ground, an NIU on the customer end and some sort of gear in the central office. This costs $285/month for on net to off net termination in my city and that is a pretty typical number.

Once you get to that termination gear you've got to negotiate the LEC's metro optical network to reach the point where they interconnect with the ISP's equipment. Despite being #53 in terms of population nationally my city doesn't have enough Sprint T1s on my ILEC to qualify for its own DS3 mux so my Sprint T1 gets drug forty miles south west to our provincial state capitol. This is non trivial, but its priced as part of that $285/month.

Once you get to the ISP's edge equipment you're probably getting 'back hauled' cross country to some location where they've got a Cisco 12000 series or some big Juniper box. You should be reading "WAN line costs", "hardended telco facilties costs", "depreciation on equipment you *can't* get at Best Buy", etc, etc.

This gets you to the ISP's network and their customers. Somewhere, out there, they peer with other top level carriers, and that is how you get to the global internet.

Besides not being able to buy the gear at Best Buy you can't *hire* the geniuses needed to make it all go from behind the counter of a local McDonalds. If you want someone who can pour piss out of a transit autonomous system without refering to the instructions printed on the heel of a Cisco 12008 you pay. If you want someone to answer the phone when the customer calls you pay. Scale that up by ten thousand T1 customers and you can imagine what is required - a real live company, so large it must be publically held to receive the funding it needs.

Bandwidth is like real estate. You can get an address on Skyline Boulevard (Sprint or UUNet DS1), you can move in to section 8 housing at 2209 Jones Street and heckle crack dealers (DSL), or maybe you're upscale enough to get a doublewide at 64th and Grover (cable modem), but make no mistake about how the world is gonna be - you plant petunias in the 'hood (VPN applications), homey's pit bull(Kazaa) is gonna take a dump there the very next day.

Re:Why an IP T1 still costs $1k/month

[rtechie]

A T1 is composed of several components, the first being the local loop to the CO. You've got two or four copper wires buried in the ground, an NIU on the customer end and some sort of gear in the central office. This costs $285/month for on net to off net termination in my city and that is a pretty typical number.

I see you failed to understand the point the user was making. In a nutshell it was:

The backbone providers are gouging ISPs on bandwith costs.

You've failed to make a significant case against this claim. The question you need to answer is:

What is the profit margin on these services?

How much money does, say, Sprint make on a $1000 a month T1? I've heard that it's almost $500. That's a 100% profit margin, which is obscene.

I don't know the detailed financials of the backbone providers, and I suspect you don't know either. Many of them are losing money, but that's mainly because of mismanagement, not because selling bandwith to ISPs isn't profitable (again, I've heard it's enormously profitable).

So what to do about it? If things stay as they are, all the ISPs will be driven out of business. And not just because of P2P. Broadband gaming (XBox, etc.), streaming media, etc. are all going to eat up more and more bandwith. And if the ISPs are forced to keep paying what they're paying, they're doomed.

I'll go out on a limb here and reccomend the obvious solution: regulation. Backbone providers are effectively public utilities and they should be regulated as such. Caps should be placed on what they can charge ISPs, etc.

Is this going to happen? Hell no. What's going to happen is that the ISPs are going to impose caps, further alienating their customers, as the ISPs continue to choke the life out of them. Eventually the vast majority will fail (mom and pop ISPs are already a dying breed) and we will see massive consolidation in the industry, with almost all the players fully or partially owned by the backbone providers.

And these megaISPs, after a few decades of screwing the consumers, will eventually eup being regulated like publich utilities anyway.

Re:Metered bandwith might help stop spam and worms

[bedouin]

IF ISPs go back to metered bandwith almost universally, they are going to be INNUNDATED with complaints that spam and getting hacked with viruses and worms are eating all the customers bandwith.

I moved my firewall to a new machine last night, and after maybe only 12 hours of letting it sit I saw that it had received 24mb of traffic on the external interface. On the internal interface, only about 4mb had been moved, which means there was 20mb's of crap hitting that machine. I suppose if I had a metered connection I'd be paying for that.