The MEC Congress starts this week as the technology reaches almost as great a peak of enthusiasm (or hype) as its ETSI stablemate, and close ally, NFV.
Rather like the 5G community – full of hopes and dreams about slicing and massive IoT, but in fact pursuing mobile and even fixed broadband use cases – MEC has tended, in real life, to leave behind its more radical possibilities and focus on a few low hanging fruits such as video caching. Yet a group of start-ups have been demonstrating a bigger vision for edge-based computing for several years now, and several of them will gather at the Congress to provide tantalizing insights into the potential of this technology.
In particular, we are looking ahead to a significant ETSI MEC proof of concept (PoC) from Saguna, Advantech, Vasona, GigaSpaces and Brocade (see below) and another PoC, focused on smart cities, from the ASTRI research institute in Hong Kong.
Their innovations will be important. It is easy to believe that MEC, like NFV, is now an accepted standard and inevitable component of the road to 5G. It is, indeed, a critical technology in terms of network efficiency and quality of service, making sense of its Cloud-RAN bedfellow. But both platforms are immature, and subject to fragmentation – this is seen in the competing approaches to NFV orchestration, respectively led by ETSI and the OpenStack community. Meanwhile, the Cisco-driven Open Fog Consortium may claim to be complementary to MEC as it also deals with placing intelligence at the very edge, especially in mobile and IoT environments, but most stakeholders are unsure how the two will work together in practice.
It is clear that flexible resource allocation between center and edge, and considerable intelligence at that edge, will be essential to make 5G economics work. The Internet of Things (IoT), with its potential millions or billions of devices, would collapse the network if every piece of data had to travel to the core. And MEC could enable far greater usage of ‘inside-out’ capacity, treating user-owned infrastructure such as home-based small cells as part of the overall RAN. Like C-RAN and MEC in general, that could be a double-edged sword for established MNOs in the same way as seamless WiFi/cellular interoperability – on the one hand, it will give them access to additional capacity at no cost; on the other, it will enable new entrants, such as the cablecos which control the indoor broadband lines, to enter their world more easily.
Key to all this is the intelligent splitting of network functions and processing, which is why C-RAN and MEC go hand-in-hand. While C-RAN centralizes core network functions as virtual machines in a private – or even (potentially) public – cloud, the risk is that it will put intolerable strain on fronthaul and core network resources, if every piece of data has to be transmitted to the core. MEC allows a great deal of processing and signalling to be done locally, often by incorporating compute and storage capacity at the cell site. When that is a small cell site, edge-based processing not only improves responsiveness and QoS to the nearby user, but can support personalization and context-aware services.
Some companies, like UK start-up and edge computing trailblazer Quortus, go further than incorporating compute power in the cell site, and actually split the packet core functions, deploying a whole virtualized evolved packet core (EPC) on a small cell or in a mobile location such as a vehicle. Such approaches turn the traditional MNO assumptions about network planning upside-down, and could enable a new set of vendors and service providers to enter the mobile market. If Quortus can implement an EPC on a Raspberry Pi or a drone (it has done both), what is the need for an OEM’s hefty EPC offering, even if that is virtualized in the cloud?
Of course, it will not be as simple as that – the key is to find the right balance between centralized and edge-based processing, just as enterprises are doing with their own cloud network and SDN implementations. Here, concepts like Cisco’s ‘fog computing’ – now enshrined in the inevitable Fog Alliance – are dominant, and there is a real need for these IT industry-driven efforts to converge and harmonize with MEC, as corporations and users increasingly go mobile-first. Virtualization and C-RAN introduce IT and data center approaches into the closed, secretive world of the mobile network – the chance for full convergence of the two worlds, rather than their current state of uneasy coexistence, must be seized if key 5G ideas like network slicing are to be possible.
The established OEMs and MNOs will try to drive this process – indeed, they must do so, because there are plenty of other organizations which have a chance to ride on mobile IT platforms to seize the rudder. These range from giants like Cisco and HPE to a host of start-ups like Altiostar and ASOCS in the C-RAN to Quortus and Saguna at the edge.
These companies feature in some of the most interesting demonstrations promised for the Munich-located MEC Congress. One, in particular, provides many pointers to how MEC will evolve. This is a MEC proof of concept (PoC) entitled ‘Multi-Service MEC Platform for Advanced Service Delivery’, in conjunction with Advantech, Brocade, Gigaspaces, Saguna and Vasona. The proposal has been accepted by the ETSI MEC Industry Specification Group (ISG).
This will demonstrate the ability to support multiple MEC platforms and applications on shared computing infrastructure, with each one adding value to passing traffic in a different way. This is a pointer towards slicing, which will allocate virtualized, dedicated slices of capacity, on-demand, to different services, with the optimal functions for that service, harnessing independent localization and scaling of the control and user planes.
The MEC PoC aims to improve on current network optimization and targeted QoS techniques such as traffic shaping, to improve subscriber experience and give operators clear, real time visibility into network characteristics and performance.
The demonstration will ride on a single compute platform supporting NFV and cloud orchestration. Traffic will be dynamically routed to the appropriate MEC platforms, where the MEC applications will provide data analytics, traffic optimization and network functions. The system leverages flexible service function chaining.
“As MEC gets standardized, it is important to demonstrate a viable MEC platform architecture, including how applications and services can utilize the platform,” said Kevin Shatzkamer, mobile CTO at Brocade, one of the enterprise networking firms, like Cisco, which could take advantage of the NFV/MEC boom.
Brocade has made a string of acquisitions in the mobile space, the most recent being carrier WiFi vendor Ruckus Wireless. Perhaps the most significant was virtualized EPC pioneer Connectem, on which Brocade’s new Virtual Core for Mobile (VCM) – targeted against Cisco’s Virtualized Packet Core (VPC) and NFV Infrastructure Solution (NFVI) -is based. VCM, announced at Mobile World Congress in February, claims to be ‘5G-ready’ to support network slicing
Brocade also positions it as an enabler of ‘packet core-as-a-service’ for hosting MVNOs in a flexible, on-demand way, or for facilitating the introduction of new services by the MNO itself. It recently launched MEC capabilities based on its vRouter, IPsec termination and firewall functions, including 3GPP standards-based MEC User Plane functionality combined with MEC APIs to link platforms, orchestration and infrastructure. These were also the result of an acquisition – 2012’s purchase of Vyatta. It says it now allows MNOs to extend the cloud to the network edge in order to support new services and revenues, and improve user experience. Another mobile-focused takeover was of Vistapointe, whose dowry was a virtual network analytics product for wireless systems.
Shatzkamer continued: “This PoC demonstrates the viability and capabilities of MEC while creating a feedback loop into the standardization process that is based on real world MEC deployments.”
The other partners are all contributing technology which has equally important indicators for the evolution of MEC and C-RAN. Vasona has been working at the edge of the network since it launched its software-based SmartAIR appliance in 2012. This sits between the RAN and the core to support mobile traffic management at the individual cell level where operators can best see how the network is behaving. It has now added a new SmartAIR Edge Services Platform, one of the platforms included in the PoC. It is a standard-based MEC software layer that runs on NFV infrastructure (NFVI).
Another platform in the PoC is Saguna’s Open-RAN, which creates and open ecosystem for third party MEC applications to run inside the RAN and as close as possible to the users, and for MNOs to deploy new services rapidly from this platform. Saguna’s heritage lies in content delivery and intelligent caching at the mobile edge, and it has several interesting alliances in this market, including one with Parallel Wireless, a small cell C-RAN start-up, and a long-standing partnership with ADVA (see inset).
Gigaspaces is providing the orchestration for the PoC demonstration via is Cloudify open source orchestrator. This onboards the components of the MEC architecture to any virtual infrastructure manager (VIM), including configuration of the various parts, using multiple protocols including Netconf and REST. Based on TOSCA, Cloudify claims to be the only open source orchestrator with multi-VIM support, for MEC service onboarding, deployment, and ongoing monitoring and maintenance all through a single pane of glass.
The actual hardware is Advantech’s PAC-6009 Packetarium XLc Carrier Grade Blade Server, based on Intel Xeon processors and targeted at locations, such as cell sites, where data center equipment would be hard to install and run.
Then there is Quortus, which was talking about MEC before it had that label and boasts of providing many of its key benefits in systems which are deployable now, even before standards are fully implemented – and even of going beyond MEC, particularly in its approach to user plane/control plane separation.
The vEPC start-up will be discussing a demonstration it first showed at the CTIA Super Mobility event earlier this month, where MEC was also a major theme. This was in partnership with Expeto, and harnesses the increasingly important technology of cloud containers with nods towards the 5G future of network slicing. Their solution will accelerate deployment of services by MNOs, MVNOs, WISPs, vertical market integrators and enterprises, say the partners – the claims which will also make slicing the hallmark of 5G.
The companies are combining Quortus’ EdgeCentrix (ECX) vEPC with Expeto’s expertise in security, containerization and orchestration, which underpin its platform-as-a-service (PaaS) offerings. This supports “IT-style” manipulation of traffic at the network edge, to boost optimization of network resources, say the partners.
ECX is now integrated into Expeto’s EPC Integrated Cloud (EPIC), a collection of containerized EPC services, plus additional virtualized network functions (VNFs) and an automation and orchestration engine. The complete solution can be deployed on-premise, in a hybrid or cloud environment, and on COTS hardware. The company says its PaaS model allows enterprises and MNOs to understand and support customers’ network demands at an extremely granular level, so they can create LTE subnets and include LTE data within their corporate security policies.
“Quortus and Expeto offer complementary technologies with genuinely disruptive potential for the entire wireless communications value chain,” said Expeto CEO Ryley MacKenzie in a statement. “With Quortus’ carrier-grade EPC as part of our solution, we can offer the best of cellular and IT-style authentication and security, quality of service and experience. We provide the entire LTE core network through a PaaS environment.”
Andy Odgers, Quortus CEO, added: “The ability to simply configure network resources at any scale, from very small to very large, and to dynamically tailor network resources to service requirements, will completely change the way public and private cellular networks are used.”
Quortus positions ECX as a step towards 5G and future enhancements to the MEC platform. It supports self-contained mobile networks for use in enterprise environments (for instance, to handle VoLTE, SIP and data locally rather than sending everything to the main mobile core) and in remote and mobile scenarios such as public safety vehicles, temporary networks or oil rigs. As small cells become mainstream, many service providers and enterprises see the potential to use this kind of approach – to deploy and control their own localized network while avoiding the cost of building a macro network by using the MNO’s platform as an anchor.
Developments like LTE in 5 GHz spectrum, and the US CBRS or 3.5 GHz band, will enable high speed in-building networks potentially controlled by non-MNO companies such as cablecos. They can use small cells and vEPCs to establish their own rules and policies. Meanwhile, even more flexibility to offer new services, with greater control over QoS than a conventional MVNO deal allows, will come when the macro network owners can provide network-as-a-service (NWaaS) on an on-demand basis, and in future, dynamic slicing. Virtualization both of the RAN and the EPC will disaggregate the network and support these new approaches, raising both risks and opportunities for MNOs (which still look to be holding onto their key weapon, licensed spectrum, in the 5G era).
Technologies like ECX can help enable the intelligent edge which is fundamental to this flexible approach. It supports ad hoc meshing, and several capabilities which Odgers sees as stepping stones to 5G, though they are deployable now – these include SDN control, the start of dynamic slicing, and critically, separation of the control plane and user plane (CUPS).
Many current MEC implementations implement the capabilities only in the user plane, especially those for video caching, the most common commercial application to date. In that scenario, all the intelligence is placed in the user plane, which creates an IT environment around the base station, in which services can be created and delivered flexibly from the cloud or locally; but leaves the network itself as a dumb pipe.
Quortus argues that the full benefits of MEC for the operator business model can only be realized with a dual-plane approach. If the network itself is left dumb, some key capabilities, such as call continuity and end-to-end security, cannot be supported at the edge, limiting the potential use cases.
Even some of the classic applications seen in user plane-only implementations can be enhanced with control plane support – for instance, in retail mall environments, any MEC system can track the numbers of users in a certain location; but with access to richer identity data via the network, services can be more accurately tailored. And a content delivery network can intelligently predict which programs a viewer will want to watch and cache them in advance.
Meanwhile, other MEC-enabled services are only possible with dual-plane functionality, including VoLTE offload, any type of push service, any service requiring session recovery, positioning-based applications, edge-based analytics and many M2M services.
Such decisions will affect operators’ future services, and so the specs which are starting to emerge from ETSI MEC and OpenFog will be closely scrutinized. In April, ETSI produced the first foundation level specs for MEC, which it says can underpin implementations in the near future, but also lay the groundwork for 5G. It released three documents to define terminology in a unified way, address technical requirements and use cases based on service provider input, and specify the initial framework and reference architecture.
The next step is for the iSG to work on nine new work items related to MEC APIs, management interfaces and essential platform functionality. It will also study MEC in an NFV environment, and work on end-to-end mobility. The former is an important element of MEC’s appeal – by tying into NFV and supporting integration of virtual network functions very close to the edge, it unifies two major trends in network deployment which might otherwise have evolved on parallel but disconnected tracks, limiting the commercial impact.
In the GS MEC 003 document, the authors write: “With respect to NFV, a mobile edge system can be realised independently from the presence of an NFV environment in the same network, or can be coexisting with it. As both MEC and NFV are based on the use of virtualization technology, MEC applications and NFV virtualised network functions can be instantiated partly or entirely over the same virtualization infrastructure.”
ADVA and Saguna demo first MEC PoC:
Saguna is part of the MEC proof of concept (PoC) which will be demonstrated in Munich (see inset), and it has also announced a demo with ADVA Optical Networking and Bezeq International.
It claims the PoC will show how MNOs can “break up vendor lock-in and offer services to business customers that could not be offered with legacy mobile infrastructures”, including those in IoT and industrial control.
The PoC 5 Enterprise Services showcase will be built on the ADVA FSP 150 ProVMe, a network edge computing platform for mobile edge application and virtual network function hosting. This combines compute resources with multiple technologies for connecting radio cell sites. This will be combined with the Saguna Open-RAN and the consolidated solution promises optimized synchronization, security and open MEC hosting.
“This PoC for enterprise services is about service efficiency, speed of service innovation and increasing the value of installed mobile networks. Removing vendor-specific barriers at the edge of a mobile network will stimulate service creativity by providing an open platform for software application developers. That’s why we’re focusing on multi-vendor PoCs like this with partners who share our commitment to open networking,” said Ulrich Kohn, director, technical marketing at ADVA.