Enhanced CDNs help to make sense of the edge compute case for MNOs

As outlined in the previous article, some of the most critical decisions mobile operators will be making as they prepare for the 5G era concern the edge.

• How reliant will their business model be on services that are focused on edge-enabled capabilities, such as very low latency or personalization?
• Will they invest in edge cloud infrastructure themselves or use a third party’s?
• How far should edge nodes be integrated or collocated with network elements like base stations or central offices?
• How can they build up a developer ecosystem, and will this involve cooperation or competition with webscale companies like Amazon, or with content providers with their edge-oriented CDNs (content-driven networks)?

Initially, the clearest case for making these decisions will relate to improving and personalizing the video experience for consumers, as this will remain the biggest driver of mobile usage for years to come. That is throwing new light on solutions which add compute capabilities and intelligence to the content delivery networks that are increasingly embedded into mobile platforms.

These decisions will help shape the services MNOs can offer and optimize on top of 5G, and how they monetize them. Some will be heavily infrastructure-focused, buiding out RAN, edge cloud and CDN assets to support high quality retail services, but also to enable a wide range of MVNOs, enterprises and, in future, service slices. Others will want to harness third party infrastructure, working with webscalers and seeking to move up the stack into device management, security, optimized content delivery and even applications.

In converging connectivity, cloud and content, and pushing all of it closer to the human or IoT user, the operators are looking at the potential to drive new revenue streams, and they are certainly creating opportunities for smart vendors to enhance their place in the MNO value chain.

Companies with software development platforms for edge-based services, for instance, have mainly been in the enterprise market, but will increasingly have the chance to cross over into the telco environment. Amazon is the most prominent example – its Greengrass platform, focused on the edge and IoT, is involved in some significant cooperations with MNO-oriented initiatives. For instance, Amazon AWS has worked with Nokia and Saguna to demonstrate Greengrass working with ETSI MEC (Multi-access Edge Computing).

This gave MNOs a hopeful sign that AWS would prefer to have the operators as partners than direct rivals, complementing its cloud platforms and services with their expertise in connectivity, mobility, massive device management and security. That balance was seen clearly in an alliance struck between AT&T and AWS last autumn, in which each partner carved out its own complementary role when selling connected cloud services to enterprises.

Other organizations are eyeing the MNO opportunity too, coming from IoT, apps development, and CDN fields. In the last category is Cloudflare, which last week unveiled a platform to allow developers to execute their code closer to end users. The solution, Cloudflare Workers, offers a middle way between processing everything centrally (with implications for latency and transport costs), and executing on the user’s own device, which can be marred by limited processing resources in the gadget, and security risks.

Cloudflare Workers is based on JavaScript, and is specifically designed to compete with Amazon’s Lambda operating environment, which was extended to the edge via Greengrass (the Greengrass software makes the Lambda environment available at the edge, either on Amazon’s own edge storage products or third party connected devices such as WiFi gateways or small cells with integrated storage). Cloudflare said its pricing would be competitive with that of Lambda, and that its platform would be accessible “within milliseconds of virtually every Internet user” via its 127 data centers worldwide. The CDN provider is expanding its presence, targeting 100 countries and 200 cities by year end, as it chases larger CDN players like Amazon CloudFront and market leader Akamai.

In the CDN race, support for edge cloud capabilities – especially to support high quality, low latency video delivery to mobile users – is becoming an important differentiator. The pairing of CDN and edge compute was pioneered by companies like Saguna early in the decade. In early 2012, for instance, Saguna worked with Jet-Stream of The Netherlands, which had deployed ‘CDN-in-a-box’, on a federated CDN solution for MNOs, which was heavily edge-based. This coincided with the shift of CDNs from wireline broadband networks – where they had been used for years to reduce the strain on the core infrastructure by optimizing multimedia delivery – towards increasingly video-driven mobile networks.

In that early example, Saguna’s RAN caching systems were combined with Jet-Stream’s VideoExchange CDN to support greater intelligence at the network edge in managing, prioritizing and delivering content to subscribers without all the data having to pass through the central core. Lior Fite, CEO of Saguna, said at the time: “Mobile operators using our joint solution can now actively participate in the over-the-top market and offer CDN services that guarantee the best possible view experience for subscribers.” The firm also made the bold claim that it could get its caching functionality into a backhaul switch or base station for less than $100.

Since then, other CDN players have intensified their efforts for mobile operators, which has inherently meant becoming more distributed and edge-focused because of the QoS issues that arise if video has to travel across long cellular links of variable bandwidth and quality. Partnerships between MNOs and OTT content providers have been enhanced by the implementation of CDNs with the operators’ networks to improve quality of experience, benefiting both sides. And the latest step is a logical one – to incorporate more processing capabilities and intelligence into content caches to support additional services and applications for MNOs and content firms.

At Mobile World Congress last month, Ericsson’s CDN business development director, Richard Dawson, demonstrated the Swedish firm’s Unified Delivery Network combined with a machine learning application hosted in a Kubernetes containerized environment. That enables Ericsson to add virtual compute resources to video caches in broadband or mobile networks. In this case, the compute power allows ML algorithms to run locally and analyze video inreal time to enhance experience or predict performance issues. It also learns to identify objects in the video stream, and so improve Ericsson’s overall ML-enabled ability to recognize images – which can then be applied to many applications. Most immediately, it is looking to overlay that information as an augmented reality layer on top of the video footage.

Ericsson has been running a trial in Sydney, Australia, to monitor webcam video of a remote robot to ensure passers-by do not interfere with its operations; and it has tested the use of its edge computing to enable synchronization of live video broadcasts with streaming video delivered to a mobile device.

In future, operators hope that they will be able to support many additional edge-focused services in low latency IoT applications, but because of the uncertain adoption paths, and monetization structures, in those markets, video is likely to be the main driver of edge infrastructure investment for some years. It certainly featured in many of the MWC demonstrations of ETSI MEC (Multi-access Edge Computing) or other edge architectures like OpenFog.

One was jointly staged by edge content delivery specialist Qwilt, in partnership with Athonet. The latter is a virtualized packet core (vEPC) start-up, which has attained high profile, thanks to its inclusion in Orange’s selection of companies to be incubated in the telco’s TIP EAC (enterprise acceleration center), under the auspices of Facebook Telecoms Infra Project.

Pointing to predictions that, by 2021, video will make up 78% of all mobile data traffic worldwide (Cisco VNI), the partners demonstrated a live mobile edge video offload system based on MEC, which supported video streaming while supporting standard 3GPP interfaces for legal intercept, security, billing and other functions.

“Mobile users are streaming more video than ever and mobile networks are not yet fully optimized to handle the deluge of live and on-demand streaming,” said Qwilt’s CEO Alon Maor. “Our edge cloud applications, designed specifically for the high demands of video delivery at the mobile edge, combined with Athonet’s Serving Gateway Local Breakout (SGW-LBO) solution, allows millions of subscribers to watch concurrent streams while maintaining a high quality of experience and optimizing network operator RAN infrastructure.” Qwilt is a founding member of the Streaming Video Alliance and the Open Caching industry initiative.

Similar demonstrations of edge content delivery combined with mobile core functionality have been developed by a wide range of large and small players, from vEPC pioneer Quortus, to CDN giant Akamai.

Last year, Akamai worked with Kaltura to put a different spin on the edge content debate. It combined its Predictive Content Delivery and CDN products, with Kaltura’s customer viewing data platform. The resulting product enables predictive on-device caching to download content in the background when the user is connected to a strong wireless signal; and with each offline download based on a user’s individual profile and history. That highlights the other great potential that arises from placing connectivity and processing close to the user – being able to deliver highly personalized and context-aware services because the mobile base station, via the SIM card, can link to the operator’s detailed profiles of its subscribers’ behaviour and preferences.

Moving to an even larger vendor, Cisco talked at MWC about its work with Reliance Jio, India’s disruptive mobile operator, to leverage MEC for optimized video experiences. They have integrated a mobile CDN into the MNO’s 4G network with edge caches in close proximity to the users, delivering content via edge cloudlets to reduce latency.

For their initial prototype, the companies used a combination of IP address management techniques based on the mobile core, implementing tight integration between edge cache, mid-tier cache, and the traffic router in the Cisco Open Media Distribution system. The mobile core and the Cisco Open Media Distribution were combined into one system-level solution for the embedded CDN. The solution implements 3GPP standard Control/User Plane Separation (CUPS) to provide greater flexibility in the choice of user planes.

In future Jio believes the same architecture will be used for low latency IoT and mission critical services, but initially, consumer video is the low hanging fruit.

Another operator which expects to invest heavily in edge resources to support consumer, enterprise and IoT services is AT&T, which recently launched its first test zone for edge applications at its AT&T Foundry innovation center in Palo Alto, California.

“We’re moving network access to cloud computation, but we’re keeping it physically close to our users. Rather than travel over wireless connections to data centers hundreds or thousands of miles away, we’ll propel this data across super-responsive 5G networks to computers just a few miles away,” wrote Melissa Arnoldi, president of technology development at AT&T, in a blog post.

The first project in the test zone centers on another low latency consumer use case, augmented reality. GridRaster provides an underlying compute and network stack to power high end AR/VR experiences on mobile devices.

Meanwhile, Deutsche Telekom will start investing in new edge computing facilities this year in Germany using technology developed by MobiledgeX, its wholly owned edge compute subsidiary based in Menlo Park, Silicon Valley.

“Part of the current plan is to take a more drastic approach and the software will be designed by a new company called MobiledgeX,” said SVP Alex Choi, who heads up network strategy for the DT group. MobiledgeX is headed Jason Hoffman, formerly head of cloud infrastructure products at Ericsson’s Silicon Valley base.

Choi, like most operator executives, will not specify how many edge nodes DT expects to deploy, although he said the operator has about 900 potential “aggregation points” in Germany alone. Some operators say thousands of nodes will be necessary per country to get services close enough to users and especially to vehicles; others say fewer than 10 will be needed, indicating the architectural differences over the edge cloud.

But Choi believes telcos have a significant advantage over webscale providers because their infrastructure is inherently more distributed and they have access to sites and to management expertise. This could be an opportunity for operators to partner with webscalers, as outlined above, rather than clash with them. Choi said in an interview: “Their infrastructure is more centralized. You can’t call one or two data centers in each region edge computing. We have great assets – towers, local exchanges and central offices – and we can place the edge computing there and make it accessible to customers.”

The German operator has already said it would like to open up MobiledgeX and make it a cross-industry initiative (perhaps with an eye on AT&T’s recent move to open source its edge software stack, Akraino). At MWC, it was reported that DT held talks with SK Telecom about potentially becoming the first outside investor in MobiledgeX. Choi may have had a hand in facilitating that discussion, since he was previously CTO at the Korean operator. The two telcos have worked together in many areas of R&D including 5G and network slicing.