Content Delivery Network Strategies, by Stef van der Ziel

By patching their networks together, many small and large network operators around the world have built a global virtual network: the Internet. It offers open and global access to any server and service in any connected network. The open and neutral character is what made the Internet the next big thing in humanity. So it is important to keep it open and neutral. And it is important to keep it running.

Iin the past few years, there has been discussion about Net Neutrality. Are ISPs allowed to block services? No? What about spam? Are ISPs allowed to shape traffic? No? Not even if a small number of subscribers can take their entire network down? Are ISP’s allowed to rent network resources to service providers so they can accelerate specific services?

In the USA, the Net Neutrality discussion is very polarized: either you are ‘freedom fighter’ who want laws to make sure that no single bit is discriminated or accelerated, or you are against government involvement and want the TelCos to be able to tune their network.

In Europe, we have a different view…

Two years ago, the Dutch government invited me to speak at a meeting about Net Neutrality. Together with TelCos and content owners we discussed the pro’s and cons, and whether Net Neutrality really was an issue. I wrote a Wikipedia article afterwards. (Dutch). I took the liberty of freely quoting and translating myself. Let’s start with the basics:

Dimensioning

Broadband access providers overbook their network. That is their business cases. They live off the margins. Consumers don’t use their link at 100% 24 hours per day. Therefore consumer grade lines are overbooked at factors between 25 and 100. Average EU households use their broadband link for 3 hours at maximum, but not at maximum speed. Therefore most consumers don’t notice the overbooking. And can pay just €20 for a 20Mbps flat-fee connection. If the ISP’s did not use overbooking, broadband connections would cost anywhere between €250 to €500 per month. We would never have achieved a 80% broadband market penetration. Consumers would stick to low bandwidth access. And internet services would never have taken off. Video services would not fly. CDNs would have no business. So dimensioning is good and important.

Top users

All broadband providers have heavy users. These users are less than five per cent of the subscribers, but account for over two thirds of the traffic load on the network. These users download illegal software, videos, films, series, games and music, all day, 24*7, mostly via Usenet or P2P sharing services. This small group claims an out of proportion part of the network resources. It has been calculated, that if these heavy users would pay accordingly to their usage, their broadband flat-fee bill would be above €350 per month. And that the broadband flat-fee bill for the 95% normal users could be as low as €5 per month, for 20Mbps! Imagine what that low price would do for broadband penetration globally. And how online services would grow.

P2P

In a previous article I wrote about Peer2Peer technologies and that they do not bring the efficiency as claimed by the developers. P2P services actually can take an entire network down. If ISP’s don’t limit P2P traffic, 95% of their customer base would be severely affected. There have been many examples where ISP’s frustrate, shape or block P2P traffic to protect their customers and their network. Blocking and frustrating of course conflicts with the open Internet idea. But shaping isn’t such a bad idea, especially if this guarantees normal use of the overall service. Many ISP’s use traffic shaping technologies and use it to do good for the vast majority of their customers.

Peering

TelCos around the world need to interconnect their networks to connect to and to become part of the Internet. TelCos can connect via transit providers, directly with physical links called private peers, or via public exchanges called public peering. Normally TelCos would have to settle for traffic that moves from one network to the other. This would have introduced such a large administrative load on every TelCo, that most TelCos started to wave costs against each other, if traffic was similar on both sides. Most TelCos try to peer privately or openly, to avoid transit costs. Both sides voluntarily exchange traffic for free. Public peering has done wonderful things for the online industry. Thanks to peering, the Internet could grow. Most traffic in Europe is delivered through open peers, we happen to sit on top of the largest Internet Exchanges in the world.

Peering abuse?

Internet service providers wanted to cut costs. Instead of paying TelCos to host and distribute their content, they setup their own networks data centers, racks or servers, or outsourced to hosting providers. And they started to make private and public peering arrangements. Large companies like Microsoft, Google and Yahoo are now publicly and privately peering with TelCos. No traffic costs. CDNs do exactly the same. In Europe, some broadcasters have built their own streaming farm and use public peering to push traffic into the TelCos networks.

A growing number of TelCos start to object to this. Because their is no mutual benefit anymore. TelCos can’t push traffic to the content providers networks. In their view, content providers, service providers and CDNs abuse peering arrangements to get a free ride on their network. The Net Neutrality principle states that if a given user pays for a certain level of internet access, and another user pays for a given level of access, that the two users should be able to connect to each other at that given rate of access. But what if the peering partner is not paying?

CDNs throw enormous loads of traffic over the fence, into TelCo networks, most of them at no charge. Subscribers pay a low monthly fee. The network load increases. TelCos can’t raise the subscribers rates for competitive reasons. Companies like Google make a lot of money. CDNs charge high costs to the content owners. The TelCos feel that their business case is under pressure while others are profiting. TelCos also want a piece of the CDN market for numerous reasons. And they have a good chance. I have heard multiple rumors of TelCos telling the CDNs to remove their servers and are re-negotiating peering agreements.

Net Neutrality abuse

Content providers, service providers and CDNs are afraid that the TelCos will abuse their position in the value chain. The TelCos control the last mile. CDNs don’t. Suppose a TelCo wants to introduce a new telephone service. They can theoretically frustrate Skype traffic to push the market towards their own solution. They can frustrate radio and tv streams to push the market towards their own IPTV service. Which would be a bad thing.

Net Neutrality rules

I am not a proponent of law or regulation regarding 100% empowered Net Neutrality. I am though for a law that forces any TelCo to pro-actively tell subscribers and service providers whether they shape, block, limit or accelerate specific services or protocols. So we came up with a set of rules:

• TelCos should always be able to dimension their network to be able to service their customers
• In order to do so, TelCos need to be able to limit excessive users and excessive or disruptive services
• TelCos are allowed to accelerate specific services or service providers as long as this does not impact other services or service providers
• TelCos are not allowed to disadvantage specific service providers in favor of other or internal service providers
• TelCos have to use the same pricing for external service providers as they do for internal services for comparable services
• TelCos have to publicly tell service providers and subscribers whether they block, limit, shape or accelerate specific services and their protocols, and have to disclose the overbooking rates

Prioritization

There has been talk about prioritization. Unfortunately most prioritization discussions are viewed upon from the TelCo standpoint. We have tried to do prioritization from a consumer point of view, and how mission critical they are:

High level critical services examples

• VoiP (requires undisturbed, low latency, realtime, 2-way communication)
• Gaming (requires undisturbed, low latency, realtime, 2-way communication)
• Streaming, IPTV (requires undisturbed, high capacity, realtime communication)
• Online transactions (require undisturbed, secure, 2-way communication)

Critical services examples

• E-mail (requires undisturbed, fast transmission)
• Surfing, searching, browsing (requires undisturbed, fast transmission)
• Chat (requires undisturbed, 2-way fast transmission)

Non-critical services examples

• Downloading (must be available, but speed is not mission critical for the consumption)
• P2P sharing (must be available, but speed is not mission critical for the consumption)

Prioritization and marketing, and transparency

Today, all TelCos offer the same service for the same price. There is not much distinction. Broadband access is a commodity. The only differentiation is price. So the price drops. TelCos are killing each other for market share by lowering their margins. That may be good in the short term, but in the long term everyone will suffer. It is time for broadband access providers to differentiate. Prioritization may actually help them. TelCos have to stop with their ’20Mbps for €20 marketing’. Prioritize! Why not focus and excel for specific markets? Many consumers would like to pay less if their ISP manages to kick out the 5% heavy users. Many consumers would like an ISP that focusses on gaming. The Gaming ISP would setup game servers for low latency. The Streaming ISP would build a large CDN to accelerate radio, tv and VOD streams from popular portals. The Social ISP would focus on P2P sharing and social services. The Pro ISP would lower the overbooking and charge a little extra for a guaranteed 24-hour service at maximum speed. And the Family ISP would squeeze P2P and introduce filters so families and their children can surf the web securely at low costs. Why not?! It’s all about choice and openness.

Acceleration!

The strange thing about the Net Neutrality discussion is that it is so negative and mainly focusses on big bad TelCos blocking, shaping and limiting traffic to cut costs or to push their own services. The discussion never focusses on acceleration which actually could be a Good Thing for service providers, TelCos, consumers, CDNs and content owners. Did you know that over 50% of the Internet traffic comes from scalable networks such as clouds, CDNs and extremely scalable hosting platforms? If there were no CDNs, the internet would collapse. So 50% of all traffic is already accelerated. Why should content owners and CDNs be allowed to accelerate content and TelCos should not be? Sounds a bit unfair, right? If you are a 100% Net Neutrality ‘freedom fighter’ you would actually be against acceleration of services. You would be against CDNs at all. And that would impact the Internet today and it’s future growth. We need acceleration, period!

Let’s look at this rule: TelCos are allowed to accelerate specific services or service providers as long as this does not impact other services or service providers

By deploying a CDN in their network, TelCos can accelerate traffic for anyone who wishes to rent these resources. If the on-net CDN is well-designed, it offloads a lot of traffic off the peers, transit links and backbones. Freeing up space for other services! 3rd party services actually benefit from other content being delivered through an on-net CDN. Content owners are free to choose to use the regular (unblocked, open) peering links. But if they want better QoS, guaranteed capacity and deeper network delivery, they can get this by paying a little extra.

CDN integration: global CDNs and on-net CDNs

On-net CDNs can further accelerate the Internet. Can enhance services. Can increase quality. Can grow your audience. Actually, TelCos with on-net CDNs can help pure-play CDNs to accelerate their business:

• The pure play CDN does not have to extend it’s network to the TelCos network, therefore cuts infrastructure costs
• The pure play CDN does not have to install and operate servers in the TelCos network, therefore cuts operational costs
• The pure play CDN can use the TelCo CDN to offload traffic deeper into it’s network than before, improving QoS
• The pure play CDN can use the TelCo CDN to offload much more traffic into it’s network than before, improving the volume
• By using a smart designed CDN, the TelCo can save costs because of offloaded transit, peering and backbone links, therefore can charge a lower cost per GB to the pure play CDN
• From a macro perspective, the overall delivery costs should be lower: Lower costs for the TelCo = lower costs for the consumer. Lower costs for the CDN = lower costs for the content owner.

It is a matter of time before TelCos will tell pure play CDNs to offload traffic via their CDN. That may be a shock to the CDN operators. They could lose control over some parts of the delivery infrastructure. But IMHO it is inevitable. It is going to happen. If the CDNs don’t follow, third parties will come to the market and work together with the TelCos and offer a better service at a lower price. Sooner or later, CDNs will have to get used to the idea to offload traffic to TelCo CDNs.

Call for inter-CDN standards

So pure play CDNs will need to integrate with TelCo CDNs. CDNs need to be able to share traffic: overflow. They need to be able to layer CDNs on top of each other: overlay. That is exactly what we have been building in our CDN technology. OverFlow and Overlay technologies allow CDNs to integrate. You can connect CDNs together and exchange traffic. You could actually arrange peering agreements on the CDN level. Hence the name ‘Video Exchange’.

Our inter-CDN technologies are built with open standards like XML and SOAP. But it is not our role to define industry standards. Therefore I am open to discuss with anyone how we can integrate with your CDN and with your CDN technology. Let me know!