On Thursday, May 21, Layer123 will host “Virtualizing Voice Infrastructure” (Register Now) a webinar from Metaswitch Networks. Among other panelists, SDxCentral Managing Editor Craig Matsumoto will be on hand, and Metaswitch General Manager Paul Drew will discuss actual network functions virtualization (NFV) deployments from his company. Drew leads Metaswitch’s Open Source Business Unit and is responsible for Project Clearwater, the company’s open source IMS Core. We talked to him recently about how NFV is changing the game for network operators and what webinar participants can expect to hear about virtualizing voice infrastructure.
SDxCentral: What will you and the other panelists discuss during the webinar on May 21?
Drew: NFV is giving network operators an opportunity to build communications networks in totally new ways. Our webinar will look at how fixed and mobile multimedia consumer communications services are changing today. We’ll give an overview of the technology behind virtualizing these services and examine how network operators are using NFV to stay competitive in a changing market.
How has the evolution of standardized commodity computing infrastructure affected the telecom industry?
Drew: Over the last decade, Intel has enhanced each successive generation of x86 processors to improve their ability to handle network-centric workloads. We’ve seen dramatic improvements in the bandwidth and packet processing throughput of x86 CPUs and rapid growth in the number of CPU cores available per unit of rack space. CPU cores now have been engineered with sizeable increases in the amount of embedded L3 cache memory and major advancements in the integrated crypto capabilities.
We’ve also seen the emergence of commodity low-cost switching silicon that can match the silicon developed in‐house by the leading router and switch vendors in terms of performance and features.
When you combine these hardware enhancements with software toolkits developed by Intel and others such as the Data Plane Development Kit, commercial off-the-shelf (COTS) servers now can support throughputs of hundreds of gigabits per second, and many tens of millions of packets per second. COTS servers may not be ready to take over the duties of core routers in the IP transport network, but they are more than capable of supporting the great majority of network functions that power today’s communications services.
What does that mean for network operators?
Drew: We believe the concepts behind NFV reached a critical mass of acceptance among Tier 1 network operators in 2012, primarily as a result of the x86 enhancements. NFV provides numerous benefits to network operators, including reductions in both capital and operating expenses and the promotion of rapid innovation in the services space.
The traditional method of building networks and services with proprietary appliances had always been expensive, slow, and cumbersome. Now, between ubiquitous Internet access and the wide availability of large-scale cloud services, software-based dot-com businesses can build IP-based communications services at huge scale and with very low cost. In some cases, these businesses have posed significant threats to telco revenues.
How is NFV changing the procurement process when it comes to deploying new services?
Drew: In the past, network operators would procure numerous pieces of equipment in order to deploy each new service. The approval process required network operators to test, certify, and purchase each piece of hardware, leading to long implementation timeframes. This stifled innovation while keeping costs unnecessarily high.
Now, with the commoditization of hardware and a progressive approach that favors network software providers (NSPs) over classic network equipment providers (NEPs), network operators can remain competitive by practically eliminating the lengthy equipment procurement processes. This dramatically reduces the time and risk for rolling out new service offerings. Using software-only solutions, procurement models can move from CapEx-weighted to almost exclusively OpEx, with subscription-based arrangements replacing large, high-risk, up-front purchases.
One of the key benefits of NFV is that it opens up the network functions market to a competitive ecosystem of NSPs. Network operators need to move away from a vendor lock-in approach, and instead engage with best-of-breed software vendors, take greater responsibility for system integration, and apply agile methodology to deliver continuous improvement.
What are some specific examples of how has NFV reduced OpEx or CapEx in actual deployments?
Drew: NFV has helped companies achieve real savings primarily through cloud management, hardware standardization, and flexible capacity
Cloud management software has automated many of the functions required to deploy a network service based on NFV. This literally enables “point-and-click” deployment of network functions. For example, to deploy a traditional physical session border controller (SBC) in the network, you would have to order a box from your SBC vendor, ship it to the location, rack it up, and connect it to power and to the physical network infrastructure. In an NFV environment, you can create a virtualized SBC in minutes entirely through simple actions performed at a cloud management console. The operational cost savings from this kind of deployment automation represent perhaps the single biggest set of benefits promised by NFV.
NFV also enables hardware standardization. In a conventional network, the failure of a piece of hardware typically doesn’t affect service, but it requires urgent action to replace the failed hardware and restore the proper level of fault tolerance. With NFV, the failure of a piece of hardware has no more affect than a temporary reduction in the maximum capacity of a given service. There is no urgency to replace the failed hardware, and the procedure for doing so is extremely simple and completely standardized.
To help network operators deal with the fluctuating load on services, NFV provides flexible capacity by enabling operators to rapidly and elastically scale their NFV deployment. This extra capacity can either be in the form of cheap, commodity, COTS hardware running a hypervisor shared with other applications with different cycles of demand, or even a public cloud. The saves on CapEx because the capacity doesn’t need to be kept in reserve using proprietary or dedicated systems that sit idle while consuming power and taking up rack space until needed.
How does NFV help operators compete more successfully with over-the-top (OTT) providers?
Drew: With NFV, network operators are able to optimize the install process in their deployments, allowing them to roll out new nodes and sites in a matter of hours or days as opposed to weeks or months for a typical lead time for a hardware-based system. This allows them to permanently add capacity in a particular location, or to roll out to a new market very rapidly, giving them the agility to compete with OTT operators wherever they see the opportunity, and to make it happen before that opportunity slips away.
Software-based network functions help drive service innovation because they enable rapid ramp-up and fast fail. With software-based functions, network operators can conduct small-scale trials of new service concepts with minimal capital cost exposure because the trials run on spare capacity provided by existing hardware resources.
This way, a new service concept that successfully gains market traction in early trials can be put into production and scaled up very quickly. On the other hand, unsuccessful experimental services can be shelved quickly without stranding hardware investment, while providing valuable lessons in what works and what doesn’t. The philosophy of prototyping new service concepts with the mantra of “fail fast, fail early, and fail often” has been a key factor in the success of many over-the‐top services, and software‐based network functions enable network operators to embrace this philosophy for the first time.
Thank you for your time today. We look forward to hearing more during the webinar.
Drew: Thank you for having me. People can register for the Virtualizing Voice Infrastructure webinar here. It will be on Thursday, May 21 at 8am PT.