Connectivity forms the backbone of numerous technologies, services, and software products that will shape life around the world in the coming decades. 5G is positioned to be a significant part of that emerging wireless landscape, as will distributed mobile computing, which takes place across connected devices.
Orange Silicon Valley analysts Manoj Mourya and Senthil Jaganathan work in this space and have a deep understanding of where the wireless world is headed. As mobile video consumption continues to play a major role in how users interact with their devices, managing that data consumption will remain an important issue for apps and networks. Manoj fielded some short questions about how he and his team see experiences and services changing as data and video services evolve, as well as what they’ve been working on and how their experiences have informed their insights.
Orange Silicon Valley: How well do telcos in general handle video over mobile networks. Is it a problem? Is there room for improvement?
Manoj Mourya: Video quality is generally a provider’s problem. For example, Netflix uses adaptive streaming, which adapts to network conditions. Many video service providers use content delivery networks (CDNs), such as Akamai, as providers for content distribution. Adaptive streaming and progressive downloads are two mechanisms adopted by video service providers to guarantee quality of experience.
Mobile network operators try to make sure that mobile core, backhaul, and radio access do not introduce much latency, jitter, and congestion to any traffic. Today, in 4G we also see lot of issues with video — the reason being air interface, because of which packet drops happen and latency is introduced. Propagation delay is another factor in bad video quality when a source of video is far away. There is room for improvement by moving a video source closer to consumer. 5G networks will offer users a maximum latency of just 4ms, down from about 20ms on LTE cells but if the video source is far away then in many cases latency will be more than 50ms.
OSV: So today, what’s my journey from my smartphone to the cloud if I want to watch a show?
MM: Today, when we watch any legal or popular website’s (Netflix, Hulu, Comcast, etc.) video on a smartphone it comes from CDN. Where is the CDN? Not in in mobile core, but rather far away. Video comes from some CDN provider (a data center) to a few points of presence (probably Equinix), then to the mobile network operator. The mobile network operator sends the video to the mobile core. Within the mobile core it has many hops (routers), backhaul, and then air interface before it is delivered to the user.
OSV: We hear about software in the network, what does that look like at the edge, and how does that help this experience?
MM: The fastest way to deliver content (video) to consumers is by placing content at the edge of the network. If the CDN is placed in the radio access network (RAN), then latency is minimal (only air interface). To place a CDN in a RAN, we need compute, memory, and storage at the RAN. We call this mobile edge compute (MEC). By RAN, I mean cloud RAN or vRAN, where a base band unit (BBU) pool is virtualized.
In our lab at Orange Silicon Valley, we placed the MEC next to a base station and did video tests on a 4G network. In the first scenario, video was fetched from Amazon Cloud, and in the second scenario the same video was fetched from the MEC. When latency (>20ms) or congestion was introduced, Amazon’s video service was not acceptable, but the video service from MEC worked fine. We have test results to show that this is the case. Amazon Web Services has operations on the West Coast. For a packet to travel from one of those locations near San Francisco to France, it takes approximately 70ms one way, which means a CDN server cannot be placed that far away; if it is, quality will be bad, because quality gets bad above 40ms. (You can ping france24.com to see round trip time.)
OSV: So why are we doing this now? What are the new enablers that make this possible?
MM: MEC is possible now because the European Telecommunications Standards Institute (ETSI) has defined specifications for MEC. Now it is referred to as multi-access edge compute. The specification is not yet complete, but some vendors have implemented it already. Compliance and interop testing will happen later. It is bit early for deployment. By the end of 2018, we are likely to see some form of deployment. In the future, some part of edge compute (mobile and wireline) will move to CORD (Central Office Re-architected as a Datacenter), but the virtualized BBU Pool will host mobile edge compute.
OSV: This feels like the next generation of content delivery networks (CDNs). Is this a new way to relate to the cloud?
MM: Telcos and mobile network operators have central offices, cloud RAN, etc. All of these properties need to be converted to datacenters (which mobile network operators own) to offer the next generation of services. These datacenters are present next to consumers, the enterprise, and industry. IoT cloud on MEC can use these datacenters for local compute. Other location-based services that can take advantage of MEC are industry automation, AR/VR, self-driving cars, drones, etc. MEC will provide standard location based API, which can be used in geo-fencing, public safety, advertising, crowd distribution applications, and elsewhere.
Mobile network operators can offer MECs as a service to Netflix, Amazon video, Akamai, etc., to put their CDN at the edge. (In short, CDN will be deployed on MEC; MEC is telco cloud at the edge.)
MEC will save cost for both the mobile network operator and video service provider. For example, 5 years back if a consumer watched Netflix on Comcast’s network, then Level 3 delivered traffic – one stream of video; if two consumers watched the same video, then Level 3 delivered two streams of video. With MEC Level 3 has to deliver only one stream of video until the base station.
OSV: Is your work on MEC local, or is there a team from HQ involved?
This activity is lead by Mr. Laurent Ruckenbusch, Head of Content Monitoring, Analytics and Mobile Optimization at Orange. We work with his team, and they are following ETSI MEC.