Using Openness to Unleash Innovation at the Network Edge Contents Accelerating Innovation at the Edge
Using OpenNESS to Unleash Innovation at the Network Edge Contents Accelerating innovation at the edge
Accelerating innovation at the edge ...... 2 The phenomenal rise of smartphones in recent Hosting applications in the network edge or in the years has shown how essential mobile applications on-premise edge has a number of advantages: Introducing OpenNESS ...... 4 have become to users. You can choose from over • Latency is dramatically reduced. The latency Simplifying orchestration ...... 6 two million apps for your phone today1, and app to the cloud could be around 100 ms, but the developers have created new revenue streams for Streamlining application reuse ...... 7 network edge can achieve latencies of 10-40 ms, mobile devices resulting in over $101 billion being A platform for innovation ...... 8 and the on-premise edge could be as fast as 5 spent in app stores in 20182. The shift to computing ms (see Figure 1). This enables applications to be Working with other toolkits and libraries ...... 9 at the edge will enable new and more compelling created that better support the high bandwidth applications, such as augmented reality (AR), virtual Use case: Facial Recognition in Retail ...... 9 of 5G. Across all network types, lower latency reality (VR) and competitive gaming. The OpenNESS advantages ...... 10 can enhance quality of service for the mobile For Communications Service Providers (CoSPs), there subscriber, and enable new latency-sensitive use is a tremendous opportunity if they can tap into the cases such as public safety applications based talent pool of mobile developers. CoSPs’ challenge is on video, and industrial Internet of Things (IIOT). to unleash a new wave of innovation at the network Entertainment and infotainment applications edge, not just in consumer applications, but also in such as location-aware gaming, content delivery business and industrial applications that will create networks, AR, and VR are also possible. new revenue streams. • Backhaul is also reduced, because some data Following network functions virtualization (NFV) will never need to go to the cloud. This can cut initiatives, general purpose compute resources can network costs, in part because it enables the now be found throughout the network and in the CoSP to upgrade network access points to meet on-premise customer equipment. These platforms increased demand without necessarily needing to could host user-facing applications, alongside the upgrade the entire network path to the cloud. virtual network functions. Where appliances have not • Data sovereignty can be enforced, for example by yet been virtualized, edge computing can be enabled enabling data to be processed in the on-premise in parallel with NFV initiatives and help to build the edge so it never leaves the data owner’s site. business case for them. This can help with sensitive or highly regulated applications.
2 However, creating edge applications introduces new challenges. In the cloud, developers are used Small Cells to a “write once, run anywhere” flexibility. In the Wireline Fixed Access Video Licensed Unlicensed (vCCAP, POXN) network, they may need to be aware of where Healthcare their app will be hosted, so they can write code to DRIVER Manufacturing Regional Data ingest data. If the edge computing resources sit FOR EDGE Centers Drones close to the base station, the data might need to be Latency Bandwidth Energy Multi-Access Edge untunneled, for example. Deeper in the network, Security Retail Computing there might be a Gateway Internet Local Area Connectivity Network (SGi-LAN) interface to contend with. Transportation Wireless Access PCs Customer Premise Some cloud service providers have capabilities Equipment Base Stations Ram Phones to host some functionality at the edge, but these (uCPE, SD-WAN, Smart Cities Edge Compute Next Gen capabilities may be relatively limited, and may tie or XXX Modes) Central Office developers in to a proprietary framework.
There is a need for a flexible, open source platform that abstracts away the complexity of the network, so that developers can write their software for the DEVICES/THINGS ON-PREMISE EDGE NETWORK EDGE NETWORK CORE DATA CENTER/CLOUD edge as easily as they write for the cloud. Latency Varies 1 ms 5 ms 10-40 ms 60 ms 100 ms expectation
Figure 1: Hosting applications closer to user devices in the on-premise edge or network edge cuts latency significantly
3 Introducing OpenNESS
The Open Network Edge Services Software Figure 2 shows an abstract representation of the (OpenNESS) is an open source reference toolkit that architecture for OpenNESS. The OpenNESS platform makes it easy to move applications from the cloud itself is shown in green. There are two parts to to the network and on-premise edge. It provides a OpenNESS: single environment for hosting applications running • The OpenNESS edge node agent runs on the in locations such as customer premises equipment edge platform. It enables data to be ingested from and small cells (on-premise edge); or the Next IP, S1_U or SGI interfaces, and supports traffic Generation Central Office (NGCO), base station, or steering at 5G latencies, multitenancy, and service wireless access aggregation point (network edge). registration. The user application sits on the same When running their applications on OpenNESS, server and uses APIs to authenticate itself and developers don’t need to think about the platform access the network, via the OpenNESS software. that will host their application, the access network The APIs are inspired by the specification for that’s being used, or the underlying compute Multi-Access Edge Computing (MEC) created by technologies. This significantly lowers the barrier the European Telecommunications Standards to entry for developers who are experienced in Institute (ETSI). cloud development, but do not have an intimate • The OpenNESS Controller can sit in the data understanding of the communications network. center, the cloud, or at the edge for orchestration of the edge platforms. The controller enables the discovery and control of connected edge devices, and is used to deploy apps to those edge platforms. The edge controller uses a web-based graphical user interface (GUI) for easy application onboarding. CoSPs can also use the APIs to link to their own controller GUI environment.
4 Currently, the supported hardware is the Intel® Xeon® Scalable processor (which can be used to bring data center performance to the edge), Intel® Xeon® ublic D processor (which offers optimized performance Edge Node loud ommercial Service Orchestrators in space and power constrained environments), Edge Native App ublic loud App loud and Intel® Movidius products (which can be used onnector A to accelerate artificial intelligence (AI) at the edge). Edge App A , E E Auth A ontroller A
Intel is working on integrating support for additional Edge ifecycle A Edge irtuali ation A accelerators including field programmable gate OpenNESS Edge Node icroservices OpenNESS Edge ontroller arrays (FPGAs), SmartNICs, and GPUs. and ata lane icroservices
Authentication and authorization of applications and OS, ontainer untime Edge irtuali ation A services are used together with the security features esource rovisionning, E A ommercial of Intel® hardware to help ensure platform security. ardware , A, Accelerators, N onfig OpenStack ubernetes nfrastructure anagers S S