Pricing CDNs

Since we’ve been talking about content delivery networks (CDNs), I did a little research to determine typical CDN services fees. We’re starting to see hybrid models, such as the GridNetworks solution streaming-media-magazine.jpg (CDN and P2P), but for strict CDN services – where content is hosted on regional servers for faster delivery – here’s what I discovered from Dan Rayburn’s recent article in Streaming Media magazine:

  • 1TB streamed: $1.50-$2.00/GB monthly
  • 5TB streamed: $.95-$1.60/GB monthly
  • 10TB streamed: $.89-$1.20/GB monthly
  • 25TB streamed: $.75-$.95/GB monthly
  • 50TB streamed: $.40-$.50/GB monthly
  • 100TB streamed: $.15-$.24/GB monthly
  • Above 100TB streamed: As low as $.12/GB monthly

These prices are nothing to sneeze at for small companies or worse, individuals. Fortunately, as MegaZone pointed out, there are other options like CoralCDN. But one has to wonder about the future of P2P given the broadband provider and customer concerns regarding bandwidth usage. Is there just not enough bandwidth to go around?

9 thoughts on “Pricing CDNs”

  1. I haven’t dealt with CDNs, but bandwidth itself is indeed pretty cheap. In a tier1 hosting facility like AT&T, you can get 1 megabit per second for between US$50-$100, depending on the time of year, your other business with them, your bargaining skills, the total volume you’re buying, etc. 1 megabit/sec ends up starting around $0.20/GB. Of course you have to provide and manage the hardware yourself to actually stream the video, and it’s not geographically segmented, so the CDNs deserve to make a profit.

  2. Bandwidth alone doesn’t solve all of the bottlenecks on the Internet.

    With more bandwidth only comes heavier load on routers, and overloaded routers drop packets. Routes advertised between routers *never* take into account quality of service. Instead, ISPs set them based upon the cheapest ($$$) route to get packets across their network. And if you are paying your ISP, and the website owner is paying his hosting provider… who is paying all of those ISPs in the middle to pass those packets between the two of you?

    Answer: nobody is. Those packets are passed through ISP peering points for free. And if you ran an ISP, wouldn’t you want to dump all of those peering packets into the slow lane and give priority to the people actually paying you?

    CDNs pay to have their servers placed in ISP data centers. This gets CDN traffic off of those peering links and into the paying customer express lanes – benefiting the CDN, the ISP, and the end-user. By the same token, it gets content closer to end-users. Everybody wins.

    Exactly who is making money off of P2P, again?

    Democracy didn’t build the Internet. Capitalism did.

  3. Jon, I agree with your points that many networks and equipment carry data. But capitalism didn’t build the Internet – and that’s exactly why we find ourselves in uncharted territory as the intricacies of a capitalist network are worked out.

  4. The modern-day Internet has, arguably, been a capitalist network since the National Science Foundation’s publicly funded NSFNet backbone was replaced by four commercially operated NAPs on April 30, 1995.

    Since that turning point, I’m of the opinion that the laws of capitalism have brought the Internet to where it is today. If companies can establish a solid business model around P2P technology, then they will survive in the end. But it will have to suffice to say that I have my doubts about P2P-as-a-business.

    Does that mean I expect to see all of the free P2P clients disappear? Hardly – I expect to see them proliferate. But I’ve come to the conclusion that when it comes to content delivery, P2P delivery mechanisms will likely always take a backseat to CDN services.

    On this, I think we agree.

  5. Having worked in the industry in one form or another since 1994, including a couple of years at GTE Internetworking, then one of the main backbone companies. (GTEI was formed with GTE acquired Internet founder BBN. GTEI got spun off as Genuity when GTE & Bell Atlantic merged to be Verizon, then Genuity went bankrupt in the crash and the remains were acquired by Level 3 – another backbone provider.)

    Peering isn’t always free – it often is with backbone providers, but not always. For example. GTEI and Sprint may have a free peering arrangement because *both* sides generate a lot of traffic *and* receive a lot of traffic, since they both have server customers and end user customers. So the traffic is generally balanced and both sides have a mutual benefit.

    The problem comes with hosting companies that do not run their own backbone. They have tons of servers, but no end users. So they generate a mountain of data flowing *out*, but only a trickle *in*. That create a massively unbalanced flow through the peering point. When I worked for GTEI we had a showdown with one of the big name hosting companies because of that. We had a peering agreement, which was free, but it had conditions on traffic, etc. They were using the free ride on our backbone to generate *massive* revenue from hosting, and GTEI wasn’t seeing a dime of that revenue. After trying to negotiate but being rebuffed, we invoked the clause and turned off their feed. That caused problems all across the net – users fed from the GTEI backbone couldn’t get to their servers, or, at best, had to go the long way around, over our network, across a peer to another backbone, and eventually to the host. Not optimal and users noticed – on both sides. So we had angry users who blamed us, and they had angry users who blamed them. It wasn’t an easy thing to do, but it costs real money to buy, or lay, fiber, put in switch gear, etc, and you can’t run a business just letting someone else use it for free.

    In the end some settlement was reached and the peering was re-enabled, I don’t remember all the details now.

    These kind of deals and standoffs happen all the time, the public rarely sees them. There is a lot of ‘horse trading’ that keeps the Internet running, and a frightening amount of personal relationships that are key to keeping things running smoothly. It is a very, very small world when you get to the backbone level, and everyone knows everyone. Personal friendships, or animosities, can play a big role in negotiations. I’ve seen first hand *extremely* critical deals that could impact millions of users settled over a couple of beers and a hand shake. There are people who work extremely hard so that we can all blog and download porn. :-)

    I also agree with Jon that, at least to a large extend, capitalism built the Internet. I’ve been online since 1989 and I remember the first commercial ISP (World.std.com), and the slow growth of the commercial presence online. But things really changed once the NSF stepped back and handed over control to commercial entities in 1995. That came just after ‘the web’ started exploding – which was the 94-95 time frame. And the commercial backbone companies, and a growing army of ISPs, poured billions of dollars into infrastructure to meet the growing demand. The nature of the net, and its reach, dramatically changed in a short period of time and growth was unlike anything that had occurred during the NSFNet days. The Internet as we know it today owes a lot to ARPANet and NSFNet – the basic protocols of course, and all the experience gained – but almost all of what we use today has been built out by commercial entities and funded by capitalist funds, not public funds. There are still bits of the old networks in there, but the growth since 1995 has been so astounding that those old networks are dwarfed by what has been built out since.

    I think the average consumer is probably better off not knowing that goes into running the backbones – just as most people are probably more comfortable NOT knowing how the electrical grid works, or the water supply, or sewers… I think people would worry more and be more nervous if they saw the guts of the operation in too much detail. You might like steak, but you may not want to tour the slaughterhouse. ;-)

  6. But do you remember when AOL created their email gateway to the Internet? Boy, did we all think our cool playground would be destroyed. And, even earlier, do you remember when MIT used to give out free shell accounts to anyone? That kept me going when I took a brief hiatus from school.

  7. Actually, now that you mention it, the AOL email gateway go-live was a pretty momentous event. I think that’s when I got my first piece of SPAM mail. There was indeed much wailing and gnashing of teeth that day. :)

  8. Dave – Of course I do. How can I forget The September That Never Ended? That was September 1993 – I even have a program that generates the date as if it were still September 1993:
    megazone @sidehack ~> ~/Bin/aoldate
    Thu Sep 5182 15:54:37 EST 1993

    I also remember when Prodigy connected to the Internet and actually stated it as “The Internet now has access to Prodigy.” Yeah, a wee bit full of themselves.

    I remember using Delphi to dial-in to WPI from NY when I was on break. Since it was designed for businesses to use it was stupidly expensive during the day – but free overnight.

    Remember BITNET? I originally had two email addresses at WPI – the ‘Internet’ address and a BITNET address.

    As for Spam – that was Canter and Siegel and the Green Card Spam.

    Ah, good times. ;-)

Comments are closed.