As senior director of marketing for Alcatel-Lucent IP Routing product group, Heidi Adams has more than 20 years of experience in the field of telecommunications and data networking. SDxCentral talked to her about how service providers are deploying carrier-grade virtualized routing functions today, and when it’s best to stick to hardware.
SDxCentral: Let’s start with the “why?” of virtualized routing: Why are network operators so interested in deploying it?
Adams: The industry is moving toward a cloud-centric approach for delivering network services, so it’s becoming more important for everyone to simplify the process from service order to service delivery. At the same time, operators still need to increase service agility so they have the ability to introduce, scale up, or scale down services as required.
Virtualized routing helps achieve these goals by enabling three main things: greater flexibility in network design, lower barriers to new service introduction, and the ability to leverage common x86-based hardware infrastructure for simplified hardware life-cycle management, operations, and maintenance
Adams: Service or edge routing is a key network function in a wide variety of use cases, and we anticipate that over time, all of these routing applications will be available as virtualized applications. Examples include the provider edge function for infrastructure routing and for the delivery of carrier Ethernet and IP VPN services; subscriber management for residential and business high-speed Internet services; and mobile packet core for 3G and LTE mobile services. Routers also are frequently deployed to support Layer 4-7 application processing for applications such as deep packet inspection (DPI) and network address translation (NAT).
In terms of when to deploy virtualized routing versus routing software integrated with specialized hardware platforms, routing functions that are typically deployed on stand-alone appliances and involve Layer 4-7 application processing are excellent candidates to be deployed in a virtualized environment. Routing applications that require a lot of processing capability, such as BGP route reflector functionality, are also well suited for virtualization.
On the other side of the spectrum, routing functions that require strict SLAs and/or require very high performance packet processing, such as aggregation or IP core routing, will be better suited to run on specialized and integrated hardware-based routers for the foreseeable future. For routing applications such as IP/MPLS provider edge (PE) routing and residential gateways, the decision to virtualize or not should be made based on the scale of the function as well as the desired target network architecture.
What are the best use cases virtualized routers can serve that traditional hardware routers would have a hard time with?
Adams: One of the key promises of NFV is lowering the economic barriers to new network-based service introduction. Today, introducing a new service to a test market involves a certain amount of up-front capital investment in networking equipment. Should the service fail to take off, there is significant lost investment. As a result, the business justification hurdle for new services is set extremely high and a lot of research must be conducted before committing to test and deliver a new service.
With NFV, the vision is that with key networking elements being deployed as virtualized network functions, there is less upfront capital investment required. Very small instances of the service can be deployed initially, and then easily scaled up should the service take off. If not, then those servers can just be loaded with other applications. NFV significantly reduces capital risk compared with traditional hardware routers in this important use case.
Do you see customers taking a hard stance on going all physical or all virtual? As software routers become optimized and general purpose CPUs improve their ability to move packets, do you see a future where hardware routers end up being obsolete?
Adams: Alcatel-Lucent’s position is to not force a choice between virtual and physical network functions, but rather to embrace the best of both and leave the choice to our customers as we work with them to make decisions on the best path forward.
Many of the world’s largest network operators are fully committed to supporting and accelerating the adoption of NFV into their networks. However, it is still early days for virtualized routing in service provider networks. We anticipate that NFV will be a longer-term journey, and that there will be a need to support hybrid networks with both virtualized routing functions and specialized router hardware-based functions for a long time to come.
We expect investment to continue in high-performance routing platforms and the specialized routing silicon that supports them. We anticipate these platforms will still be required to address continuing growth in bandwidth demand in the most economic way possible.
What criteria should network operators use in evaluating and selecting virtualized routers?
Adams: Network operators should seek software that meets three main criteria. First, it should be based on proven software that has been deployed in both small-scale and large-scale operations in a real-world environment.
Second, the routing software should be designed and optimized to fully leverage cloud compute infrastructure. Generally this is indicated by support for a 64-bit software architecture, as well as support for a multi-core processing environment and parallel task execution at scale.
Finally, the routing software should be architected for deployment in either integrated or fully distributed fashion. This allows different functions to be shared across different virtual machines, and it indicates that the routing software can take advantage of cloud scale-up and scale-out benefits.
As virtualized routers take hold and large numbers of them get deployed, how do you see network operators managing them all?
Adams: Whether they’re deployed on general purpose platforms or on specialized router platforms, both deployment models will still require configuration and provisioning, monitoring, and fault management. Element and network management functions do not go away.
As you move to NFV, there is the added requirement of managing and automating the underlying NFV infrastructure, as well as the applications that reside on top of it. This is where an NFV management and orchestration (MANO) system designed for handling carrier-grade applications becomes key. Alcatel-Lucent’s 5620 SAM management system operates is a seamless fashion across both virtualized and integrated router instances.
Finally, let’s hit the all-important question of “When?” When do you think virtualized routing will hit mainstream, and which use cases are getting the most traction right now?
Adams: Overall, there is still a journey ahead of us before virtualized routing in service provider networks will be considered a mainstream deployment option. Initial testing, evaluation, and implementation initiatives are well underway today and have focused on router functions including mobile gateways, virtualized route reflection, and router simulators. The provider edge function, and customer premise equipment applications – for example, residential gateway or customer premise IP routers – are also seeing strong interest and are expected to be the next applications to move to the deployment phase.
SDxCentral: Thank you for joining us today!
Adams: Thank you for having me.