Your browser is not supported. Please update it.

9 October 2014

MNOs seek to convert cache into cash

Some MNOs (Mobile Network Operators) are planning to offer in-network caching services to CDNs and also content providers in an attempt to hook some revenue from the proliferation of OTT. The MNO factor has risen rapidly up the CDN agenda over the last year as witnessed at conferences such as last week’s CDN Summit in London, which featured a number of Cellcos. The main focus was on the role of mobile CDNs to cope with the traffic deluge caused primarily by OTT video such as YouTube, either as an alternative or complement to WiFi offload. But there are also signs of MNOs looking to cash in on edge cache by connecting directly to CDNs or even sources of content, cutting out the ISP in the middle that has until now usually provided the link.

This is interesting in so far that up till now MNOs have been reluctant to open up to CDN providers and allow deployment of edge servers within their  core networks, partly through absence of a viable revenue sharing model. Another factor is that external parties might disrupt their network operation and also bring security risks. Typically therefore CDN edge servers have been located outside the MNO’s domain. Then a control application executes a distribution algorithm to select which edge server should store a copy of certain content to optimize the quality of experience (QoE) for end-users. But with this caching process outside the MNO’s control, delivering end to end QoS for video or for that matter, m-commerce has proved elusive. At the CDN Summit John Hakansson, Head of CDN at Telia Sonera, Europe’s sixth largest Telco, emphasized the need to focus on web site load times for OTT and commerce services. “If that takes five or six 6 seconds, customers are leaving you,” he said. For Telia Sonera therefore CDN strategy is as much about improving QoE and combatting churn as deriving new revenues.

Japan’s KDDI Corp is set on turning cache into a cash cow. The decision to start deploying cache inside its core network was initially driven by the consolidation taking place within the CDN industry itself, according to KDDI’s manager of IP Transport Network Development for CDN Technology Kaxuhide Koide. “A large proportion of traffic is now flowing into our network from a handful of CDN operators,” said Koide. “So we have to consider cooperating with these operators to improve quality and performance and develop the technology to improve accessibility of OTT.” There was also the proliferation of encrypted traffic via the HTTPS security protocol which is most efficiently and safely handled inside the core network, but another factor that has loomed larger with time is the potential of caching in the network both for cost saving and revenue gain.

“Traditionally we have an ISP for transit between the CDN and MNO,” said Koide. “But now we are looking to cut out the ISP and finally have the CDN paying for cache and additional CDN services inside the MNO network.” KDDI would also be targeting OTT content providers, especially those such as Netflix that are building their CDNs in any case.

This resonates with the corresponding situation for fixed line broadband networks, which are also being connected with CDNs, especially in the US. Here too, separate ISPs have widely provided “middle mile” connections between the CDNs and the broadband access networks and according to data from Netflix, these were themselves often the bottlenecks constraining the end user’s bandwidth and QoS. Otherwise it was the interface between the ISP network and the last mile broadband network that caused the problem. In either case the issue is being resolved for Netflix by direct connection of its CDN, called Open Connect, and the broadband access network.

But Netflix still complained about having to pay for such direct connection, with Comcast for example, on grounds of network neutrality. It would be interesting to see if the same argument will be made for direct connection to MNO networks, given that here too Netflix can argue the operator benefits through improved QoS for its users, as well as possible bandwidth savings.

Koide is convinced that either CDN or content owners will willingly pay for cache so long as the MNO delivers added value. His point is that over mobile networks QoS cannot yet be taken for granted and therefore is something that content providers and ultimately consumers will have to pay for, at least for now.

There remain some issues to be faced. One, acknowledged by Koide, is that common interfaces are needed to connect MNO networks with CDNs to lower deployment costs. Then there is the issue of “cacheability”, given that there is no benefit in storing content near the edge that is only requested by a single user. Cacheability is defined as the ratio between the amount of content revisited and the total amount requested. It is a crude measure because it does not take account of number of revisits, but on that basis cacheability is usually about 50% measured by numbers of hits but only 10% in volume. The discrepancy is because content accessed a lot, on YouTube for example, is more likely to be consumed in small bites with a high dropout rate. Recent analysis of YouTube video and other user generated content (UGC) sites showed that on average only about 10% of each item is watched. The value of caching such content is therefore less than it might seem.

But if these factors are taken into account caching within the network looks like it will at the very least save money for MNOs and possibly create new revenue as well.