Network virtualization is the overarching concept for the move to NFV and telco cloud. This evolution hasn’t been easy. Poor performance, high cost and lack of interoperability have all been impediments. Now we’re seeing breakthroughs that are knocking down yesterday’s barriers. In this blog, I describe some requirements for success in this evolution. I also include some success stories about real-world deployments.
It Starts With Product Features
Cool features aren’t sufficient for success, but they are necessary. Here are some essential capabilities for enabling network virtualization. Good news for service providers – these features now exist and are available at prices that meet tight budgets.
- Networking. Service providers are in the business of providing networking services. Virtualization solutions can’t be limited to the data center; they have to extend out into the network. Service providers need tools to virtualize today’s services. Moving from hardware appliances to software network functions is a start, and requires support at multiple protocol levels. Even more importantly, service providers need ways to innovate in network services.
- Performance and Cost. Network virtualization means deploying servers out in the network. Individual installations of servers mean controlling costs is critical. At the same time, service providers need high performance to ensure carrier-grade services. In the past, high performance has meant high cost. This conflict between performance and cost has now been solved. Recently we have seen reductions in processor cost and increases in processor performance. At the same time, we have seen improved software efficiency. Together, these improvements are good news. Service providers can meet the requirements for service offerings AND the business case.
- Application Scalability – Up and Down. One of the benefits of software-centric cloud technologies is the idea of “write once, run many.” A single software application can run in a multitude of environments, scaling up and down as needed. This notion is very valuable for service providers. In the past, service providers have tested and certified individual products for each application. Now, service providers can deploy a common application for multiple environments – from residential to small business, to headquarters, to data centers, all with one solution.
- Universal CPE. The flip side of application scalability is a uniform hosting environment. This means a standard software environment hosted on a standard server, even when deployed at the customer site. This application is now being labeled as universal CPE. uCPE enables service providers to deploy commercial off-the-shelf (COTS) servers at a customer site. They can use these COTS servers to extend their cloud-native hosting environment – from the data center all the way to the customer site.
The changes needed for network virtualization start with features like those listed above, but that’s only the start. The next question is: How do you deploy network virtualization at scale? Here’s how.
- Embedded Cloud. OpenStack is an essential element of today’s data center deployments. But service providers have tripped over some issues when moving OpenStack from the data center to the telco network. These issues include scalability, upgradeability, and security. Now there is a way to realize the benefits of OpenStack while resolving the issues. We call this model “embedded cloud.” With embedded cloud, we run an OpenStack controller in a container on a uCPE platform. This makes the uCPE a miniature standalone cloud or a “cloud in a box.”
- Zero Touch Commissioning. Service providers have been deploying networking appliances for a long time. The result is features that are optimized for simplicity in deployment. One of these features is zero touch commissioning (ZTC). With ZTC, an appliance is installed without the need for specific configuration. Instead, the appliance “calls home” to a server to get its configuration. In the past, hosting environments running on COTS servers did not have this capability. Now they do. A service provider can have a COTS server drop-shipped to a customer, and use zero touch features to turn up virtualized services. Service activation will no longer take weeks and months – it can be done in days or even hours.
- Manageability. Service providers have a long and rich experience with operating networks. They have high expectations for how virtualized networks should be managed. These expectations include service turn-up, monitoring, maintenance, troubleshooting and decommissioning. These expectations drive requirements. Virtualization solutions must have open APIs for integration into service provider management systems. In addition, the suppliers of these solutions must provide support and integration services.
- Security. Service providers want the benefits of cloud technologies. But they are concerned about security issues. Virtualized solutions must be designed for security from the ground up. The layering of defenses, ease of upgrade and visibility are all essential elements.
Moving to cloud technologies means more than software features and processor cost. It means changing behavior.
- Multi-Vendor Ecosystems. Service providers have previously purchased closed solutions from a single supplier. This approach simplified operation and support, but it put the service providers at the mercy of the suppliers. It also limited innovation. Virtualized solutions must be built from components from multiple suppliers. Service providers must work with hardware and software suppliers to make this approach succeed.
- Urgency and Focus. Traditional telco solutions had a long lifespan in the network. Service providers worked hard to ensure they picked the right solution. “Resolve everything before doing anything” was the order of the day. Now, service providers want to move rapidly. They want to develop solutions that are mostly right, deploy them quickly, and use the experience to iterate and improve. SD-WAN on uCPE is a great candidate for starting on this path. Customer want it, and it can be added as an overlay to existing networks.
- Organization, Agility, and DevOps. The speed described above requires its own set of changes. Moving quickly means breaking down organizational barriers. It means new ways of working, such as Agile development. It also means using DevOps for mechanizing operations.
We Are Seeing Success With Deployments – Get on Board!
The move to network virtualization is not theoretical. It’s happening now. Here are some success stories:
- Masergy. Masergy was the first service provider to deploy virtualized services on a plain server at the customer site. The Masergy Virtual f(n) service replaced a stack of physical devices with software running on a server. This customer deployment was coupled with a sophisticated management system. Masergy worked closely with their suppliers to solve the problems of cost and performance. The result: Masergy customers could use a portal to dynamically order services.
- CenturyLink. CenturyLink has been driving network virtualization for several years. They have committed to replacing network appliances with software, and have shown remarkable progress. This includes worldwide deployment of virtualization pods, support for CORD and progress in virtual infrastructure.
- Verizon. Verizon is a leader in Network Virtualization. Verizon’s Virtual Networking Service (VNS) was a first step towards a cloud native framework for service innovation. Verizon has now added to that with the universal customer premises equipment (uCPE) program. uCPE takes the VNS vision and pushes it to the customer site. In addition, uCPE adds some valuable features such as using wireless connectivity to accelerate deployments.