CDN Federation: A Badly-Defined Solution in Search of a Real Problem?
At last year's Content Delivery Summit Europe, there was serious interest in exploring the CDN federation model. Further investigation, however, reveals that the notion was more down to marketing and sales than engineering
At last year's Content Delivery Summit Europe there was a general consensus that if a CDN federation model was to ever be more than a discussion topic then some action would help it along.
We gave the topic the closing hour of the day, and we committed to supporting it in the programming on an annual basis. There was a willingness to progress the discussion and participate in a trial among a number of operators, and even the offer of support from a potential sponsor.
My plan had been to talk with one particular CDN leader who in my mind undeniably had a good model for CDN federation, and then to start a working project from a base provided by the sponsor.
On talking to him, and the potential sponsor this week, some four or five months later, there is a sense that the market has changed, and that CDN federation is fraught with complexities that perhaps were not evident in the autumn
The interesting thing for me has been that while there is a broad will to engage with CDN federation, I now believe that this stems from a sales and marketing perspective, rather than an engineering perspective. In effect I’m sensing that the longer running and more established a CDN, the less interest there is in a federation and this leads to some observations.
The largest CDNs have little interest in federation—they are more interested in bringing their own infrastructure footprint "into" third-party networks than establishing a working process that lets the third party a) potentially derive revenue from the deepest points in the network, and b) gives the third party visibility of their clients and traffic volumes. For a globally expanding, bullish CDN this is tantamount to enabling their competition. It's counterintuitive.
While there are parallels with mobile virtual network operation (MVNO) in the mobile telecoms market, there are also huge differences: The CDN typically has a few valuable sources (content providers) and many destinations. The MVNO has many sources of small value and many destinations. Visibility of the clients of one MVNO by its competition doesn’t really change the status quo: There are a LOT of mobile phone users on any MVNO, so roaming doesn’t really represent a competitive threat. However in the CDN Federation model for a global CDN providing a regional telco the insight that you are delivering terabytes of content provider A’s video to hundreds of thousands of the regional telcos’ clients sets up the telco with information that enables the regional telco to ring up content provider A, take a direct feed of that content and undercut the global CDN.
On the other hand, it comes as no surprise that smaller regional upstart CDNs looking for new clientele (and these could be small death-star or regional overlay CDNs) are keen to federate since they have the most to gain in terms of new revenue and market intelligence. Given they are typically of the scale that they can maintain a close developed relationship with their own clients, the risk of losing a client is reduced, too.
Also, one has to ask if it isn’t just simpler to increase the peering capacity and let the CDN commercial landscape continue to evolve along fairly established rules. At the end of the day, capacity provisioning is much simpler than deploying a new federation and interoperation strategy.
It’s been a while since I hit a problem with streaming that was down to insufficient core network and long haul issues; interference with my Wi-Fi because popcorn is being made in the microwave in the kitchen is frankly far more likely. These days CDN is a core efficiency and cost-economic, and has little to do with end-user experience, and so while on the outside of the box CDN federation is about making an ever-more congested internet less congested, this is not really the case: Yes, it is true that at the very edges into the domestic and enterprise environment proper proxy caching can greatly improvce the end users experience, the upstream issue of delivering Tbps into an ISP is relatively solved. CDN federation shouldn’t worry too much about that.
At the end of the day, it was always going to be more focussed on interchange of commercial information—such as rights, revenue collection, and subscriber management—than it was truly to do with ensuring distribution. However while any CDN wants to see what traffic a competitor is shipping, this just doesn’t seem to be information that CDNs really, in their hearts, want to provide to their competition.
My conclusion is that CDN federation is currently very much a "badly defined solution looking for a real problem."
That said we still have a sponsor interested in backing a federation trial, and I'm still keen to be proven wrong: If there are players out there who see a CDN federation working and would like to add the group who are interested from last year's Content Delivery Summit and perhaps the support of a sponsor, then please do get in touch on firstname.lastname@example.org.
Attendees of the Video Infrastructure Summit in London will attempt to push federation forward, and also discuss online entertainment and privacy.