OpenStack is becoming the de facto standard for infrastructure orchestration for NFV deployment by leading Communications Service Providers (CSPs). CSPs are trading off the challenges of OpenStack implementations (e.g. immature technology and evolving standards) for the benefits of open source and open architectures (i.e. reduced vendor lock-in). Lack of standards for NFV management and orchestration (MANO) remains a leading impediment.
NFV and OpenStack
OpenStack is a set of open source software tools for building and managing cloud computing platforms. It enables service providers to provision and orchestrate pools of data center resources across compute and storage. With regards to NFV, CSPs deploy different types of data centers than the typical large enterprise or hyper-scale cloud provides. Their compute capabilities are distributed in tiers including centralized cores, aggregation points, and local points of presence (PoPs). OpenStack implementations for CSPs must be highly reliable and be able to distribute workloads across hundreds of geographically distributed data centers (of varying sizes).
Current Implementations of OpenStack in Leading CSPs
Recent announcements by leading CSPs have shown significant support for OpenStack as part of NFV deployments. According to the OpenStack Foundation, CSPs are the fastest growth sector for OpenStack deployments. OpenStack is a key element of many emerging NFV standards including: CORD (Central Office Redesigned as a Data Center) and OPNFV.
Verizon (working with Big Switch Networks, Dell and Red Hat) has detailed OpenStack cloud deployment across five of its U.S. data centers. Its NFV project, which began in 2015, relies on a core and pod architecture that provides the hyperscale capabilities and flexibility necessary to meet the company’s complex network requirements.
Additional deployments are in progress in additional domestic data center and aggregation sites, with international locations to be deployed during 2016. Verizon also plans deploy modified data center designs at edge network. According to Verizon, it plans to virtualize a large portion of its wireline and wireless networks elements (using NFV and OpenStack) over the next 3 years –with the scale of deployment including tens of thousands of servers.
AT&T plans to virtualize 75% of its network over the next few years by leveraging NFV, SDN, OpenStack, OpenDayLight, and Open vSwitch. AT&T’s NFV cloud already includes 10 OpenStack projects with three more planned by the end of 2016. Leveraging open source code (including OpenStack) AT&T’s goal is to reduce deployment times for cloud resources from months to days.
AT&T has done a lot of work to modify OpenStack, including extensions to improve its reliability (carrier grade), to enable its use of part of the foundation for its new global network. AT&T is also using OpenStack tools to develop an end-user resource manager. This will give AT&T customers a self-service capability to enable network on demand services.
Other capabilities that AT&T has enabled with NFV and OpenStack include an application catalog, which allows for automated provisioning of network functions and a Region Discovery Service, a reservation system enabling applications to run across various geographic regions on the network.
Other large CSPs with public support for and implementation of OpenStack include: NTT, SK Telecom, China Mobile, China Unicom, XO Communications, Deutsche Telecom, Swisscom, British Telecom, Telefonica, Comcast, and Time Warner Cable.
Benefits of OpenStack for NFV
Leading CSP are looking for many of the same benefits as other large data center deployments when deploying OpenStack. They hope to lower both deployment (CAPEX) and operating costs (OPEX) as well as improving the flexibility and pace of innovation for their cloud-based resources. Large CSPs specifically mention their desire to eliminate vendor lock-in as a key factor behind their interest in OpenStack. Other factors include the power of innovation inherent in open source projects, broad ISV ecosystem, and open APIs.
Challenges of OpenStack for CSPs
OpenStack is a cloud management system, not a generalized network management system. This means for CSPs that OpenStack is just one piece of the management and orchestration puzzle – but not in itself the solution. For example, OpenStack lacks the mechanisms to support many of the services and service components now deployed using legacy network equipment.
Some of the weaknesses of OpenStack are likely to be fixed as the technology matures and becomes widely deployed including:
- Ability to scale to thousands of sites and hundreds of thousands of servers
- Consistency in deployment of various OpenStack components
- Lack of comprehensive documentation
Other challenges will be harder to implement and may require CSP specific implementations, including:
- Security of network data – especially east-west traffic between data centers
- Ability to tie OpenStack to the CSP’s broader NFV management and orchestration solution
- Multi-vendor VNF resource allocation and support for service chaining
Wide range of suppliers support OpenStack as part of NFV developments, including: HPE, DELL, VMWare, Cisco, Red Hat, Mirantis, Ericsson, Big Switch, Pluribus Networks, PlumGrid, Intel/Wind River, Radisys, and many others.
Recommendations for CSPs
OpenStack is now a de facto component of NFV deployments as the means to provision data center resources. It is a piece of the complex NFV puzzle that allows CSPs to leverage open source innovation and reduce vendor lock-in. OpenStack will require enhancements in terms of reliability, geographic scale, and security to meet the specific requirements of leading CSP NFV deployments. OpenStack does not solve the tremendously complex challenge of developing standards for NFV management and orchestration (MANO).