Peering directly is not enough to ensure peak performance

07 November 2017 | Mike Hollyman

Cover

Mike Hollyman

Blog Author | Nokia Deepfield; Head of Consulting Engineering


Content from cloud applications, especially OTT video traffic, is increasingly dominating the internet.

To manage the burgeoning demand, ISPs are utilizing new techniques to ensure delivery of the highest quality streaming to subscribers. To keep their customers satisfied, they must, among other things, efficiently manage peering relationships. This requires them to understand exactly what the traffic is, how it is being delivered and from where it is being requested. 

For instance, for an ISP to deliver optimal quality video from a popular content provider, such as HBO, it must build complex peering relationships with HBO’s contracted CDN, Level3. These relationships ensure the links provide equal coverage across all areas of the ISP’s network. But, what happens if HBO suddenly changes its business model to deliver content from its own CDN? The ISP will need to quickly develop peering relationships and provision bandwidth to this new CDN in order to deliver the same volume of HBO traffic (while also factoring for growth). If the ISP fails to recognize and plan for these network shifts, the network might still handle normal traffic but buckle during the release of popular programming such as HBO’s Game of Thrones. 

In other words, failure to recognise and plan for network shifts in peering relationships may cause a significant decrease in video streaming quality and result in a flood of customer support calls. The bottom line is that an operator needs tools to see exactly how much traffic is entering into the network. These tools can help operators understand the kind of traffic and through which peering points it is entering the network. This enables the ISP to either add additional capacity to a peering interface or shift overflow to transit links at peak. Both options have a cost, but having the insight and data to make the best decision is absolutely essential for long-term network health and revenue.

This isn’t just a fanciful example. Dropbox provides a real life example of a cloud application that migrated CDNs in this way. When Dropbox was just emerging, it used Amazon Simple Cloud Storage Service (S3) to distribute its content. However, as the number of users of Drobox increased, it built its own CDN and began transitioning users away from S3. In order for ISPs to have ensured optimal traffic delivery as capacity grew to Dropbox’s CDN, it would have been very helpful to have had a tool that understood how much Dropbox traffic existed on the network when it was delivered by S3 – and then be able to provision sufficient bandwidth for the new peering relationships with the Dropbox CDN.

Visibility into the source and passage of all traffic allows networks to be managed more efficiently. ISPs need a multi-dimensional, holistic view that combines complex data from the cloud and the network. Simply put, they need visibility of what is flowing to and through their networks. This holistic view should also enable them to follow encrypted IP flows, identify the origin and destination of all traffic related to subscriber content, and uncover traffic traversing less than ideal peering connections. Such a multi-dimensional perspective on both cloud and network traffic data can fully arm the ISP to monitor traffic growth and plan for future capacity increases. Ultimately, it is a critical piece for optimising the customer experience.

About the author

Mike Hollyman has 25 years of experience as a network engineer and security professional, working for large enterprises and service providers, architecting highly scalable solutions to meet the needs of today's next generation networks.