Mobile backhaul networks have had to bear the brunt of performance and capacity improvements at the RAN level at least since 3G started to boost data rates and especially with 4G LTE. Now 5G is ratcheting the pressure up by another order of magnitude.
The industry’s research community has been working on further optimization improvements for some years now. The fruits of such endeavors are starting to emerge and be ready for deployment alongside 5G roll-outs. Operators now understand clearly, in the wake of their LTE experiences, that mobile backhaul cannot be considered in isolation but should be planned and optimized in concert with enhancements to the whole end-to-end infrastructure from the core to the RAN. Changes in cell site antennas, for example, will affect the best arrangement of cache hierarchies in the backhaul.
It is worth remembering that an effective caching strategy itself should operate end-to-end and not be confined to the backhaul. It begins at content servers in the core and ends on the mobile devices themselves, which are increasingly used for pre-emptive caching where pieces of video content might be downloaded at quieter times like overnight. This would be done when there is a significant probability the user might want to play that content later at a time the network is congested.
We also have to consider the emerging portion of the infrastructure called fronthaul, which began as intermediate fiber links between radio heads in the RAN and centralized radio controllers, at the front of the traditional backhaul connecting to the core. This concept has expanded to include complex topologies linking large numbers of radio units to virtualized basebands in the cloud; aerial hubs such as drones; and tethered balloons that have been considered as an intermediate layer for small cell deployment in dense urban areas, connecting to the core network directly or via a traditional backhaul.
It makes sense to include such fronthaul options in the bigger backhaul picture, since they are together addressing the common challenges associated with 5G but actually being driven by user trends and technological advances across the board. These trends are primarily proliferation in capacity and in the number of radios associated with 5G deployment, along with downward pressure on latency associated in particular with live streaming video but also conferencing applications (themselves boosted by the COVID-19 crisis).
Caching inevitably lies at the heart of backhaul optimization, whether for fixed-line or mobile services. It dates back over half a century to the 1960s when it was introduced to accelerate access to large data stores by moving the most commonly requested data into a unit with less capacity but faster read times. This was much later applied to the Internet for caching popular web pages in smaller distributed servers closer to the user rather than a single central server, trading storage for network bandwidth while reducing latency.
But as web traffic grew further in the late 1990s such ad hoc caching arrangements became increasingly inefficient, unwieldy and hard to manage and the CDN (content delivery network) was born. CDNs quickly expanded as dedicated vessels for organizing data transport and associated caching over large distances, leaving ISPs just to relay the data locally to the end user. It was left to the CDN to optimize long distance data transmission by locating servers judiciously, allocating the right storage capacity to each one, deciding how and where to store content according to its demand profile and working out how most efficiently to route content from the caches out of their domain, which itself moved increasingly close to the end user.
CDNs carried more and more video, firstly on demand but then increasingly live as well, foisting new complexities. Video generated a lot of data but at varying dimensions because of the variety of encodes used that meant a stream of given content might range from 128Kbps for low quality mobile services to 30Mbps for ultra-HD to large smart TVs. Behavior of live video especially is also very different, with one advantage being that because of its timing constraints it is at least predictable when streaming requests will arrive, which helps operators optimize their caching and routing for that traffic.
Meanwhile mobile services had brought their own caching demands with the desire to trade precious wireless bandwidth and transmission power fort relatively cheap hard or solid state disk storage, which can be used to stage live transmissions so that they can be broadcast over the RAN. Then just a single channel is needed to reach multiple users within a cell. The other big trend relating to mobile backhaul is the rise in edge computing, to reduce fixed bandwidth and latency at the same time.
Various vendors in the CDN space have come out with optimizations for mobile backhaul taking particular account of video in the context of 5G. All of these are still work in progress and further improvements in mobile backhaul optimization will continue to emerge as 5G roll-outs progress and as results from research programs are incorporated.
UK-based CDN technology developer Velocix is the latest to commercialize the fruits of research with software addressing three technical aspects of cache operation within backhauls and CDNs. Although presented as specific for mobile backhaul, the enhancements would be equally welcome to networks whose last mile goes over fixed infrastructures, but 5G provides a strong focal point for marketing and promotion.
Furthermore, 5G presents some specific challenges to the backhaul around which Velocix has optimized its new software, coming from the RAN and the core. On the RAN front there is NR (New Radio) bringing a big jump in bit-rates by using wider channels, enabled by the higher spectral range available in the millimeter wave bands. Those wider channels and also the smaller cells required to cater for the reduced range at these high frequencies both affect the backhaul.
There are also changes in the core to service the greater density and lower latency. More particularly, both the core and the RAN, and therefore also the backhaul in between, must cater for 5G’s network slicing feature, where the whole end-to end physical network can be “virtualized” or sub-divided into logical partitions dedicated to given applications, services or use cases.
This has one higher level motivation, to facilitate sharing of costly infrastructure between multiple operators, rather than just MVNOs. Slicing also enables partitions of a network to be dedicated to given use cases with different requirements, such as video conferencing needing ultra-low latency, ultra-HD streaming where the need is for consistent delivery at high bit rate, and many IoT applications where the need is to support large numbers of often very small data transmissions. Backhaul networks need to underpin the slices and associated use cases.
Velocix’s three main components for mobile backhaul optimization are cache layer bypassing, inter-cache communication and multi-access edge computing (MEC), which between them increase bandwidth efficiency and reduce latency. While backhaul networks might be all-fiber in an ideal world, in practice this will not be attained at least for now and so there is pressure for mechanisms that alleviate bandwidth over slower copper and microwave links, while also helping 5G services meet their much more ambitious latency targets.
CDNs and backhaul networks are usually arranged in hierarchies of caches radiating out from the center or core through which long tail content flows like a waterfall over each layer. More popular content is cached more locally and so does not traverse as many cache layers. This process has become less efficient as content has proliferated, such that a large amount in total is lowlier rated for popularity and therefore involves requests that have to filter to a central server up this cache mountain.
This consumes a lot of upstream bandwidth which therefore has to be provisioned at the expense of downstream capacity that could serve content. The Velocix cache bypass allows upstream requests to jump straight through the backhaul network, often provided by a CDN supplier, to the server where that content resides. Velocix claims this can save 30% of upstream bandwidth and while there is no independent verification that figure is quite believable, if the emphasis is on the word ‘can’.
More to the point is how much upstream bandwidth it would save on average for a typical 5G backhaul network and that is not yet clear. Velocix says that the figure was derived from modelling performance against a 5G reference architecture incorporating statistics from real world deployments and insist that a 30% reduction from cache bypass is achievable for the larger operators typical of its customers.
However, Velocix also offers inter-cache communication to improve upstream efficiency with a promise of even greater bandwidth savings in that direction of 50%, in this case based on modelling a typical multi-tier CDN reference architecture. If the two figures were cumulative, that would amount to 65% saving in total.
The inter-cache communication is achieved through more effective sharing of content at the network edge than in the traditional CDN approach where popular content is balanced across multiple servers in a point of presence (PoP). When a given user requests content that is spread across say two edge servers and one of them is busy at the time, the transaction is handled by the server that is still free. That server does not have all the content and cannot obtain it immediately from its peer on the edge and so has to go back up the line towards the origin to get it. Velocix now allows this content to be partitioned between edge servers such that when one of them gets close to being fully occupied its peers can butt and pull the content off directly. This will reduce latency, cut consumption of upstream bandwidth and also alleviate pressure on origin servers.
The third enhancement is MEC at the periphery of the backhaul networks, moving into the fronthaul territory, sometimes called edge clouds. These may be located between the operator’s backhaul or internal IP network and the 5G RAN in switching offices. These edge clouds can be virtualized in the sense that they run on commodity hardware that may be shared with other applications under the banner of ‘edge computing’ and Velocix is presenting this as part of its complete package for scaling up to 5G backhaul and meeting the much higher anticipated traffic demands. Velocix points out that MEC promises to reduce the amount of rack space and power consumed in data centers by standardizing on a single, multi-tenant edge framework for apps.
This will not be enough on its own and further innovations will come along as the caching ecosystem is in as great a state of flux as it has ever been, driven partly by this explosion in mobile occasioned by video in particular. Such innovations are being deployed by the major CDN providers such as Akamai, Limelight and Level 3, as well as Ericsson’s Edge Gravity rebranded from Unified Delivery Network (UDN), in partnership with about 80 telcos.