All In with Virtualization? An Argument against Complete Virtualization
Everyone is talking about how the virtualization of the network and services infrastructure is going to revolutionize the network architectures of today. I am hearing how the entire network will move from proprietary hardware, such as routers and switches for SDN and they are also discussing how services like firewalls, services gateways, and core service provider network components such as the IMS infrastructure and packet gateways within the NFV framework.
As these virtualization technologies are fairly new, they are being investigated for all aspects of the current network architectures. Some parties envision a fully virtualized network that replaces the need for proprietary hardware with common off-the-shelf (COTS) hardware utilizing an X86 architecture. In reality though, it is hard to develop a manageable roadmap and plan to create a virtualized network that is able to perform with the same capabilities of the existing networks.
There Ain’t No Such Thing As A Free Lunch
Virtualization allows one to leverage the benefits of agility and elasticity found in cloud architectures with properly embedded management and orchestration systems. But this agility and elasticity comes with a price. Additional operational overhead and reduced performance introduce additional complexities into the architecture.
COTS hardware does not have the same high performance characteristics as proprietary hardware. This means more instances of the function are necessary to deliver the same network performance capabilities. Multiple instances need to be aggregated using ADC and load balancing technologies. This is potentially an additional function that needs to be incorporated for successful virtualization technology execution.
Another layer of infrastructure necessary to support the virtualized architecture is the need to incorporate a hypervisor management system to enable the implementation of these virtualized services on the COTS hardware. This adds to the complexity of the operational management of the virtualized infrastructure.
When to Call and When to Raise
These issues lead to two conclusions. First, it is important to identify which services it makes sense to virtualize. Because of scaling and the necessity to deliver high network performance services to deliver functions like packet data network firewall where aggregate throughputs can exceed 40 Gbps, it makes sense to retain the high performance proprietary hardware models that can handle the capacity of these areas of the network architecture.
Second, architects need to look for key critical components and functions where operators can take advantage of the agility and elasticity of the delivery of these services to the end user. This usually means services that have variable demand requirements and need the resiliency of geographic redundancy. In the service provider environment, this typically starts at the value-added services (VAS) and then extends into the packet data network (SGi in an LTE mobile environment) and then the rest of the core service provider network.
Emerging technologies create a vision depicting the impact of a disruptive technology. Remember that with any new technology and its evolution it is important to remember that practical implications of the revolutionary technology as it relates to the existing network architecture models.