The Need for Experience-Driven Backhaul
As new applications evolve with the continued development of wireless technology, they change the experience expectations of consumers. To ensure that these shifting experience thresholds are met, it's essential that total network performance is aligned with the characteristics of current and future apps.
With HDTV, 2K and 4K Ultra HD TV, the critical capability is throughput. Experience on video is also related to the evolving adaptive bitrate HTTP streaming technologies, such as Apple’s Live Streaming (HLS) and the international standardized solution MPEG Dynamic Adaptive Streaming over HTTP (MPEG-DASH).
Each of these adaptive streaming solutions enables the client's device to adjust the bit rate on a per segment basis, dependent on available bandwidth and measured by network throughput and delay. The intent is to select and download the segment with the highest bit rate for play back without stalls or re-buffering. With MPEG-DASH, which has flexible segments, the network needs to ensure consistent high throughput to ensure a one- to two-second initial waiting time and the best experience.
In a one- to two-second adaptive streaming world, backhaul performance needs to be monitored at the level of seconds, possibly even tens of milliseconds, rather than every five, 10 or 15 minutes. It also needs to be measured on a per-service, per-user basis. This requires a rethink of backhaul architecture from the days of 2G/3G, where Web traffic was merged with voice traffic, to today's real world of LTE and LTE-Advanced, where Mobile TV and OTT services are demanding ever more throughput.
As the network as a whole is being upgraded to handle greater bandwidth, and using technology such as small cells, carrier aggregation, cloud RAN and coordinated multi-point operations, it is necessary to include backhaul as an integral part of the upgrade.
A key question should be what are the technology options to bridge, from a network centric perspective of backhaul performance to a customer centric one of quality of experience? Increasingly, network planners with an eye to the future are recognizing that, beyond throughput, latency and synchronization are the critical requirements for future success and survivability.
This is particularly the case for real-time applications, such as VoLTE and, over the next five years, vehicle-to-vehicle communications. In these cases, sub-optimal engineering could deliver poor latency in one or both directions of a link, resulting in a totally unacceptable user experience, or worse, a crash. In this new world, one way latency matters as much as round-trip latency, which creates more challenging engineering requirements.
Increasingly, operators’ roadmaps include Cloud RAN (C-RAN) with fronthaul architectures. The concept has the baseband pooled in centralized or cloud servers, and links to separate RF heads via fiber or wireless over a standard interface, at distances of up to 15 miles. Whatever the architecture, the cell sites need scalable backhaul capacity that is phase-synchronized with the rest of the network, to ensure throughput of both burst and microburst traffic.
Backhaul phase synchronization is required to support LTE Broadcast and for LTE network capacity features, such as CoMP and eICIC, that have much more stringent synchronization requirements. Phase synchronization restricts the deviation in the exact moment when a cell site clock can tick, relative to the Primary Reference Clock (PRC). Because there can't be any buffering or averaging, attaining these stringent targets requires significantly raising backhaul performance to eliminate or reduce the risk of jitter and delay impacts.
In order to get down to these required synchronization levels, operators need to think in terms of eliminating tens of nanoseconds from the delay budget. This requires considering a distributed timing mechanism, using concepts such as Precision Time Protocol (PTP) mini grand master clocks closer to network edge, supported by boundary clocks and even transparent clocks to dynamically update the PTP packet. This is not a simple exercise since some older equipment can't be upgraded and there are variations in vendor implementations.
As the network densifies with HetNet and smaller cells, there is increased pressure for network feature distribution closer to the edge, and the use of aggregation or hub sites. These hub sites can be used from a backhaul perspective to support the CoMP feature that requires X2 handover latency of less than a few milliseconds, and which is better achieved in a hub site than the core.
This distributed capability also plays to the introduction of network functions virtualization (NFV) and software-defined network (SDN) paradigms into the backhaul architecture. The way that backhaul networks operate today, network nodes typically have little or no visibility into conditions on other links in the network that could serve as an alternative path. With SDN, the controller has link visibility and control of all network elements and links in software, which means that optimal paths can be calculated, which can improve latency.
In order to achieve the end-user video and voice experience thresholds that consumers will increasingly expect, or to be able to charge OTT providers for premium delivery options, it requires a close coupling and integration of backhaul with the future network enhancements. It also demands a significant rethink of the current mobile network architecture regarding the balance between the need for intelligence in the core and at the edge, as well as the potential use of SDN to ensure optimum routing of traffic under all conditions going forward.
— Steve Bell, Senior Analyst, Heavy Reading
This blog is sponsored by Huawei.