By Deng Ao
With the proliferation of LTE, basing VoLTE on the IMS has become a carrier priority. IMS features high performance, high integration, and high-capacity computing. It proves the value of telecommunications network virtualization and is required for NFV.
At MWC 2014, network function virtualization (NFV) was one of the cutting-edge technologies that drew the most attention. It promises a beautiful blueprint for carriers, but NFV realization is no easy task. Carriers have to find more NFV application scenarios based on network situations and propose practical evolution solutions.
NFV: Revolutionizing network architecture
In January 2013, 13 top companies in the ICT industry jointly established the ETSI NFV ISG work group to define NFV requirements, architecture, and application scenarios. So far, nearly 200 carriers and venders have joined the work group.
NFV aims to decouple software from hardware and virtualizes network functions so that they do not have to depend on special hardware, and new functions and services can be deployed through software installation and upgrade. Hardware can also be upgraded at the pace of Moore's Law, enhancing network performance and cutting cost. NFV helps carriers increase network flexibility, realize highly efficient network construction and operation, expedite service provisioning, and reduce total cost of ownership (TCO).
NFV is an important breakthrough for carriers. It also represents the next network architecture revolution, following the transformation to all-IP.
IMS: Ideal breakthrough for NFV
During the early stages of research into NFV, a universal platform is required that can adapt to all network functions. However, this revolution in network architecture cannot be achieved overnight, due to numerous difficulties related to power consumption, efficiency, virtualization of real-time services, and reliability. If we divide the network into service, control, and user planes, we can find the breakthrough points for NFV by analyzing the difficulties and benefits of their virtualization.
The service plane provides value-added services (VAS), and some of them run on a standard hardware platform, similar to an IT server. The service plane is easy to virtualize and the benefits are obvious. However, its virtualization does not touch upon the key problems of telecommunication network virtualization, such as maintaining the SLA after cloudification. Virtualizing the service plane could be a first step even if it contributes little to the overall network virtualization process.
The control plane is responsible for the scheduling of network resources. It requires real-time signal processing capability and high reliability. High computing capacity is the major need of control plane virtualization, which can be satisfied by a universal IT hardware platform. Thanks to mature hardware, control plane virtualization can address many deep-rooted problems such as carrier-level SLA guarantee on a universal hardware-based network topology, and NFV network management & maintenance. Therefore, virtualization of the control plane would pave the way for overall network cloudification.
To meet requirement of high-throughput packet forwarding capability, high reliability, and low latency, traditional platform adopted many private hardware such as NP processors, speed-up ASICs, and hard codecs. The current universal IT platform has yet to meet the demands of network IO, and transcoding. For example, the signal processing of the wireless carrier system consumes too many CPU resources, causing high latency and low throughput. What's more, the aggregation of various user-plane functions is also challenging. In summary, the virtualization of the user plane still has a long way to go.
Our analysis shows that virtualizing the control plane first is the most feasible choice. The current universal hardware meets the demands of control plane virtualization with regard to service reliability, platform virtualization, and centralized resource management, while the control plane itself is the network brain and key to transformation from network-based to service-based networking. LTE proliferation makes commercialization of IMS-based VoLTE a priority. VoLTE and NFV began to show promise at nearly the same time, giving IMS virtualization a higher priority in control plane virtualization. According to Infonetics' SDN and NFV Strategies: Global Service Provider Survey released in July 2013 and ABI's The SDN and NFV Business Case released in September 2013, IMS will be among the first elements of the telecommunications network to be virtualized. Leading carriers such as Deutsche Telecom (DT), Vodafone and Telefónica are already conducting POC and testing NFV in VoLTE deployment. Huawei is their preferred partner in this journey.
vIMS Realization: Key problems facing
Global carriers are sparing no effort to promote NFV maturation and deploy VoLTE based on virtual IMS (vIMS) during core network evolution to LTE. So what are the challenges of vIMS implementation?
The NFV ISG is responsible for formulating the framework for NFV, and is developing NFV-requirement whitepapers for standardization organizations such as 3GPP, IETF, and ITU-T. In May 2014, the sixth NFV meeting started to review group specifications developed by work groups, and discussed the target & plans of NFV Phase 2 (which will be started in the first quarter of 2015). Therefore, NFV now has a stable framework, but it will take time for it to mature.
SLA and flexibility
In the IT field, services are HTTP-based and not real-time. Carrier services are SIP-based, real-time and reliable. Therefore, it might degrade the SLA when directly migrating real-time carrier services to IT platforms. To guarantee the accessibility and real-time feature of carrier services on a distributed Internet-based framework, designs and optimizations must be in accordance with the requirements of carrier service models and every single network function.
E2E integration and network O&M
Using special devices for VoLTE deployment is complex and involves over 40 NEs and 60 interfaces. IMS-based VoLTE deployment requires multiple layers of entity systems (hardware, virtual software, and application software) to be reintegrated, so as to ensure highly reliable operation of IMS on the virtual platform. Then integrations and IOTs of NEs with standard interfaces are needed. This is called vertical integration and it poses many challenges. The clarification of interfaces between multiple layers of entity systems, compatibility, and unified & effective network fault allocation methods must be addressed. What's more, different service venders have different SLAs for carriers, leading to variation in response time and unclear responsibility between vendors.
NFV friendly O&M
Carriers have developed mature network management and operations & maintenance (O&M) systems in the past decades. When the network is revolutionized by NFV, the existing management and O&M must be upgraded at the same time. Carriers' organizational structure, for instance, needs to transform from vertical to layered. A synergized management and maintenance for both traditional and NFV networks is needed. Associating alarms with virtualized network functions and virtualized hardware position and type is also an issue. Another challenge lies in establishing a unified network fault location method and an end-to-end (E2E) fault detection & recovering mechanism between multiple heterogeneous network layers.
In the future, the evolution target of maintenance systems will be to realize unified dashboard-based management, automated service scheduling, application installation, service provisioning and self-healing of network faults over distributed data centers, heterogeneous virtualized services, and interoperation between different cloud platforms.
CloudIMS: Forging IT-based carrier networks
Based on advanced research and a profound understanding of the ICT industry, Huawei has proposed its future-oriented SoftCOM strategy. Based on NFV and SDN, future network architecture will be cloudified, making it more dynamic, efficient, scalable, automated, and open. A cloudified network supports cloud management, cloud services, cloud control, and a cloud edge.
To realize the virtualization of the control plane for SoftCOM, Huawei announced its core network virtualization plan, which aims to virtualize all core network devices, including IMS, PCRF, CS, and USC. Huawei commercially launched its virtualized IMS solution, Huawei CloudIMS, at MWC 2014, which supports large scale automated deployment, elastic expansion, and virtualized network resources. In addition, vIMS guarantees the SLA of carrier services after cloudification.
Huawei actively contributes to the development of NFV. Huawei is a co-chair of INF WG and plays an important role in NFV development. Huawei hosted the second NFV conference. By the second quarter of 2014, Huawei submitted 207 proposals, more than any other member. Huawei is a gold member of OpenStack and is devoted to open source development of NFV.
Under the guidance of NFV visions and objectives, Huawei CloudIMS will be opened and decoupled step by step. FusionEngine is ICT convergent off-the-shelf (COTS) hardware. It can be managed by third parties through SNMP and IMPI and passed IT-standard certification and testing, including VMware authentication and SPECvirt testing. It can be powered by AC and deployed in the IT equipment room. FusionEngine meets various strict telecommunications requirements, including ETSI and NEBS Level 2 certification, eight-degree shock strength, and has a working temperature range of -5 to 55°C.
Huawei offers an OpenStack-based open cloud platform – FusionSphere, which supports over 200 types of hardware (computing, storage, and switching), major virtualization technologies and guest OS. It can carry both IT and CT applications and provide improved performance that is both reliable and real-time.
In terms of virtual network functions (VNF), all IMS functions including SBC are virtualized. Huawei CloudIMS is an open solution that is compatible with Huawei NFV infrastructure (NFVI) and other forms of NFVI, such as VMware and HP COTS.
Realizing IT flexibility and high SLA for telecommunication services
FusionEngine is a carrier-enhanced COTS platform that effectively reduces the impact of virtualization on real time performance. FusionEngine uses a variety of advanced technologies in the chip layer, including the VT-X-supporting CPU, SR-IOV-supporting network interface chip, and 100G switching chip that realizes microsecond-level latency. In system architecture, FusionEngine realizes unified computing, storage, and switching. The computing and storage capacity are one of the most powerful in the IT industry. 400G per slot high bandwidth is also supported. Built-in eFabric supports flat Layer-1 network topology, which isolates north-south traffic from east-west traffic and minimizes external latency caused by multi-level forwarding and switching.
FusionSphere is specifically optimized to support real-time telecommunication services. Advanced DPDK technology is used and powerful switching capacity is aligned with single root I/O virtualization (SR-IOV). What's more, telecommunication affinity scheduling, large memory pages, interruption aggregation, the embedded operations interruption (EOI), along with the advanced programmable interrupt controller (APIC) all help reduce latency for service processing. In service assurance and a reliability, live migration of VM and a fault tolerance (FT) mechanism ensure quick fault recovery.
NEs for vIMS use cloud-aware architecture, realizing statelessness and decentralization and maximizing perception-based service flexibility by the isolation of applications and data. More importantly, process-level N+M redundancy, in combination with the inter-DC pooling of VNFs, ensures quick service recovery and high service availability.
High-efficiency automated network deployment and service scheduling
Based on NFV ISG's recommended architecture, management of NFV networks is accomplished by the management and orchestration (MANO) system. MANO consists of three layers. Each MANO layer is responsible for resource management and scheduling on one of the three network layers. The first layer is the virtualized infrastructure manager (VIM). The second layer is the virtualized network function manager (VNFM), which manages the lifecycle of virtualized network functions (VNFs). Huawei's VNFM can be independently deployed or deployed as an expansion module of U2000 (EMS) to minimize impacts on the BSS and OSS on existing networks. The third layer is the network function virtualization orchestrator (NFVO) that manages the lifecycle of network services and conducts service scheduling. The southbound interfaces of Huawei's NFVO can be used for managing third-party VNFM. Its northbound interfaces work with a large data analytics system to implement accurate lifecycle management of network services. All three layers can interoperate and coordinate with external modules through the REST interface. The coordination and association of the three layers make vIMS-based network deployment as easy as installing computer software. What's more, by coordinating with existing EMS/OSS, MANO can inherit functions such as KPI, alarm association, hardware management, and application alarm, realizing E2E network O&M monitoring.
Easy to deploy
vIMS applications can run on Huawei's open-cloud platform or third-party platforms. Carriers can customize their networks to suit their specific situations.
Networks built on Huawei's NFVI components (FusionEngine and FusionSphere) and vIMS have the following advantages:
Simplified integration: Standards are currently incomplete, but through E2E pre-integration, pre-verification, and unified service, the problem of vertical integration is solved, avoiding the potential risks of multi-layer NE integration. This will also shorten TTM and quickly meet service requirements.
Future-orientation: Networks built on Huawei's NFVI components (FusionEngine and FusionSphere) and vIMS are a foundation for TOC and ICT transformation. As the industry standards and NFV mature, any third-party COTS can be added to the current hardware layer to provide computing and storage capabilities. Third-party applications can also be deployed on the current platform to enable sharing of NFVI by multiple services.
Another option is to deploy networks using Huawei's vIMS and third-party NFVI (VMware + HP COTS). In 2013, Vodafone conducted POC verification of Huawei's VMware-based vIMS solution. Huawei's solution leads the industry in terms of deployment time and testing cases. Huawei also established a comprehensive integration service framework and supporting teams to help carriers tackle challenges by providing primary integration service. Through comprehensive network design of the application, network, and system layers, Huawei can guarantee high SLA standards for telecommunication services on platforms of low reliability. By constructing the capability of unified maintenance of heterogeneous components from different vendors, Huawei can help carriers realize easy IMS deployment, high availability/reliability, and simple network O&M in multi-vendor scenarios.
Huawei is actively promoting the development of virtualization for the telecommunications industry and hopes to cooperate with carriers to explore integration, unified O&M capability construction in multi-vendor scenarios, SLA guarantee for telecommunication services on NFV networks, and virtual network management. Ultimately, IT-based telecommunication networks can be constructed. That is the ultimate goal of NFV.