Anwer Al-Dulaimi, EXFO, Canada, Shahid Mumtaz, Instituto de Telecomunicações, Portugal and Qiang Ni, Lancaster University, UK
Published: 30 Apr 2018
CTN Issue: April 2018
A note from the editor:
This month we consider some of the challenges that face the deployment of 5G. In particular we look at deployment on cloud based networks and the challenges of network slicing 5G in this environment. Anwer, Shahid and Qiang provide us with a comprehensive picture of the pieces that will have to fall into place before we can get to the promised land of a truly software defined and sliced 5G network. As always, your comments are most welcome.
Sliced 5G Services in Cloud-based Networks
The fifth generation (5G) networks will employ cloud-based architectures in their deployments. This means that baseband processing units and core network entities will be virtualized and instantiated at various network clouds in a hierarchical order. Therefore, 5G virtual appliances will be cloud native applications running at various datacenters in a clustered network orientation. The concept is to process clusters traffic demands locally and align with other clouds for intra-cluster calls or when demanding additional computational resources. Although this operational scheme seems to be logical and cost efficient, sliced 5G service type models require us to rethink the distribution of resources between cloud segments. In this article, we show the association between distributed clouds and radio interfaces for better architecting interworking between physical and virtual components.
- 5G Service Slicing
5G networks are integrating different segments and technologies to work as one complete unit for service delivery. Therefore, the radio resources are not the only limiting factor for operators’ capacity expansion. In fact, networks may not be able to fully utilize transmission opportunities due to end-to-end (E2E) service processing cycle. Regular network operations include processing requests, prioritizing services, establishing links, accessing application servers, enforcing charging schemes, and more. Network densification brings network access points closer to users for better connectivity and spectrum savings. Similarly, the concept of bringing baseband processing closer to users is continuing to evolve through architectural modifications to network structure. For instance, mobile edge computing (MEC)  and network function virtualization (NFV)  are key enabling technologies of local service processing in softwarable networks. Moreover, software defined networking (SDN)  and artificial intelligence (AI)  support smart traffic diversion between various entities’ virtual and physical ports. From an architectural perspective, operators are laying down the basic infrastructure for cloud-based ultra-dense network (UDN). However, the 5G developers need to confirm the theoretical requirements (e.g. latency and capacity) in large testbeds and filed trails that employ E2E 5G techniques.
Baseband processing capabilities and traffic assignment in service-based architecture are the other dimensions that network designers need to consider when reaffirming performance expectations for 5G networks. Processing higher load requests with different cores speeds add additional latencies to current service processing causing overhead for computational and radio interface components. In other words, the computational resources may not be compatible with radio resources or vice-versa. This is more likely to occur in clustered networks that deploy geographical distributed datacenters (geo-datacenter) inside network domains. The 5G is a find-grained network that drive slicing across physical and virtual resources to maximize the amount of services for consumers. Each 5G Slice/Service type (SST) specifies the latency metrics for enhanced Mobile Broadband (eMBB) supporting virtual reality/augmented reality (VR/AR), ultra-reliable low latency communications (uRLLC) including enhanced Vehicle to Everything (eV2X), and massive Internet of Things (mIoT) , as shown in Fig. 1. Regardless of specific numbers, different SSTs demands latency to be in the order of single digit milliseconds or less. This many raise questions on how to architect underlaying infrastructure to support intensive 5G latency requirements.
The 5G Core Network (5GC) will be deployed as software packages that run in cloudcenters. Therefore, 5GC entities and baseband processing units (BBUs) will be all deployed as virtual machines (VMs) but in separated silos. Virtual BBUs will be located at MEC to process local radio access network (RAN) waveforms . While 5GC will be deployed at various local and national datacenters to manage calls in a hierarchical model. In a virtual environment, the orchestrator triggers automated service creation subject to changes in arrival traffic. Therefore, VMs are launched or terminated using a life cycle management framework that speculates network state transitions. Various network state transitions may cause additional time delays for processed traffic when scaling VMs or mapping resources.
The VM is running on a hypervisor that abstracts virtual resources from the computational resources of a commercial-off-the-shelf (COTS) server. All these hardware and software components require processing time to perform their functionality. These systematic time delays are still incorporated even with orchestrated services where all network operations are fully automated. For instance, the time required for elasticity while scaling an application demands resource allocations and reconnecting virtual to physical networks. Although this time is very minimum and may be managed in seconds, it will certainly impact the overall E2E estimates for any type of 5G services. There are two more issues to consider: hypervisors competence and hardware specifications. Hypervisors continue to evolve with new features and engines that support more automation and better scalability. Moreover, new plugins are added to support 3rd party software including VMs.
5G networks will use a combination of operators and 3rd party clouds making it harder to maintain consistent performance due to diversity in platforms features. The networking is also another consideration when transferring the traffic between virtual networks and physical network interface controllers (NICs). There are multiple types of such bridging that differ in their use cases and advantages for users (e.g. vswitches, data plane development kit (dpdk), single root input/output virtualization (SR-IOV), etc.). It is not clear what is the time required to process packets and distribute flows in complex networks. For UDN, this impose additional time delays in each segment of a typical 5G transport network. Obviously, the network performance will be affected because of those latencies incorporated at different networking technologies. This will help characterizing performance of underlaying platforms including COTS and virtualization software. Such analysis helps to decide the computability of basic network components considering various 5G slice type requirements.
Blending RAN and Cloud
In 5G transport network, the different segments require conversions from wireless type packets to ethernet and/or fiber frames . In large-sized networks, packets need to travel throughout various network segments and probably virtualization platforms. In such mixed networking, the SDN provides another layer to overlay connections and reduce the time latency between originating and destination nodes. However, the dynamic updating of routing tables is crucial to the success of such systems and any failure in editing flows may incur significant delays Therefore, it is necessary to consider the performance of network platform (hardware and software) as a service before being able to support reliable and scalable sliced 5G services. The current assumptions of sliced network performance need verification before being able to confirm compatibility with 5G requirements such as eV2X. To this end, we think that network control modules should have the ability to assign incoming service to proper network platforms that can meet performance metrics for that service type. This does not require operators to deploy various types of hardware and software alongside. However, operators need to leverage key enabling technologies in the form of orchestrated service catalogs to utilize available resources. In this way, RAN and 5GC are interconnected using predefined networking schemes that are chosen with the arrival SST. Finally, RAN and clouds need to be blended using telemetry services for real time monitoring of E2E network status and control.
- C. M. Huang, M. S. Chiang, D. T. Dao, W. L. Su, S. Xu and H. Zhou, "V2V Data Offloading for Cellular Network Based on the Software Defined Network (SDN) Inside Mobile Edge Computing (MEC) Architecture," in IEEE Access, vol. 6, pp. 17741-17755, 2018.
- A. Basta, A. Blenk, K. Hoffmann, H. J. Morper, M. Hoffmann and W. Kellerer, "Towards a Cost Optimal Design for a 5G Mobile Core Network Based on SDN and NFV," in IEEE Transactions on Network and Service Management, vol. 14, no. 4, pp. 1061-1075, Dec. 2017.
- S. Al-Rubaye and J. Aulin, "Grid Modernization Enabled by SDN Controllers: Leveraging Interoperability for Accessing Unlicensed Band," in IEEE Wireless Communications, vol. 24, no. 5, pp. 60-67, October 2017.
- M. Yousif, "Intelligence in the Cloud – We Need a Lot of it," in IEEE Cloud Computing, vol. 4, no. 6, pp. 4-6, November/December 2017.
- A. Prasad, A. Benjebbour, O. Bulakci, K. I. Pedersen, N. K. Pratas and M. Mezzavilla, "Agile Radio Resource Management Techniques for 5G New Radio," in IEEE Communications Magazine, vol. 55, no. 6, pp. 62-63, 2017.
- C. Sexton, Q. Bodinier, A. Farhang, N. Marchetti, F. Bader and L. A. DaSilva, "Enabling Asynchronous Machine-Type D2D Communication Using Multiple Waveforms in 5G," in IEEE Internet of Things Journal, vol. 5, no. 2, pp. 1307-1322, April 2018.
- K. Nakajima, T. Matsui, K. Saito, T. Sakamoto and N. Araki, "Multi-Core Fiber Technology: Next Generation Optical Communication Strategy," in IEEE Communications Standards Magazine, vol. 1, no. 3, pp. 38-45, SEPTEMBER 2017.
Anwer Al-Dulaimi is a System Engineering Specialist in the R&D department at EXFO, Toronto, Canada. He received his Ph.D. degree in electrical and electronic engineering from Brunel University, London, UK in 2012. He is the editor of IEEE 5G Initiative Series in IEEE Vehicular Technology Magazine, associate editor of IEEE Communication Magazine, and editor of vehicular networking series in IEEE Communication Standards Magazine. He is the chair of IEEE 1932.1 Working Group “Standard for Licensed/Unlicensed Spectrum Interoperability in Wireless Mobile Network”.
Shahid Mumtaz is a Senior Research Scientist and Technical Manager at Instituto de Telecomunicações Aveiro, Portugal. He received his MSc and Ph.D. degrees in Electrical & Electronic Engineering from Blekinge Institute of Technology (BTH) Karlskrona, Sweden and University of Aveiro, Portugal in 2006 and 2011, respectively. Dr. Mumtaz has more than 90 publications in international conferences, journal papers, and book chapters. He is serving as a Vice-Chair of IEEE 5G Standardization. He is also an editor of three books and served as an editor for many IEEE journal calls.
Qiang Ni is a Professor and Head of the Communication Systems Research Group, School of Computing and Communications, Lancaster University, UK. He received his Ph.D. degree in 1999 from Huazhong University of Science and Technology, China. His research interests include 5G, Green Communications, Ultra-Dense Networks, Cognitive Radio, SDN, UAV, IoT, Smart City and VANETs. He is a Voting Member of IEEE 1932.1 standard. He was an IEEE 802.11 Standard Voting Member and contributor to various wireless standards.
Statements and opinions given in a work published by the IEEE or the IEEE Communications Society are the expressions of the author(s). Responsibility for the content of published articles rests upon the authors(s), not IEEE nor the IEEE Communications Society.