At Procera we are working passionately and persistently on answering the big questions of NFV. Questions such as what is the lynchpin of NFV? How fast can NFV go on multiple machines? Without asking the right questions the answer is as meaningless as the classic "42" from the The Hitchhiker's Guide to the Galaxy.
Running network appliances virtualized is a quite straight forward story given that you can get to the necessary performance and efficiency, the real lynchpin of NFV. But is performance good enough for multiple virtual machines? For certain use-cases, it is not enough to have great performance for just one virtual machine, you need to have great performance for a multitude of virtual machines. Procera has shown that we can obtain great performance on multiple virtual machines. We proved this by running 150 Gbps over a single COTS Intel Haswell server with multiple virtual machines using SR-IOV The Intel XL710 NIC allows for 64 virtual functions (VF) allowing for stunning network performance for 64 virtual machines per physical NIC. A great step forward from the earlier PCI pass through of one high performance VM per physical NIC.
Can virtualization open up new business cases? For certain business cases using appliances simply is not possible as it is too expensive. A good example of this is the distributed vCPE case.
For Boingo Wireless, a classical appliance approach simply could not work. Airport regulations and a staggering amount of sites across the world would require specialized hardware competence in 90 countries for tens of network functions – an environment that would make it too expensive for Boingo to operate their WiFi products with specialized appliances. In this case, virtualization, leveraging Procera’s PacketLogic/V and VMWare, was the only way to go to make the business case hold. Try it out yourself by testing a Boingo WiFi network at the next airport you visit.
So far I’ve written about virtualization. Let me also spend some time on NFV by asking “What is different about the ESTI NFV architecture over classic data center virtualization?”
In NFV, virtualization is the foundation with many other important parts that are required, such as the Virtualization Infrastructure, SDN, Orchestration and Infrastructure Management. This allows for innovation at each layer of the stack by different companies who are experts in their field.
A great example of this is when companies work together to create the best-of-breed solutions. In show of the great collaboration occurring in this new space, Procera is working on the ETSI NFV POC#32 together with Vodafone, Openet, Red Hat, Intel and Amartus to evolve the next generation vEPC.
In the use-case of centralized vCPE for enterprise, the need for orchestration becomes very obvious. On-boarding a new enterprise customer when you have hundreds if not thousands of potential customers needs to be a quick and fail-safe process with minimal human effort. My background in the OSS/BSS space with focus on consumer broadband last mile provisioning has shown me the great challenges with having thousands of new customer activations per day. It must be automated. Running a POC for a Tier-1 service provider in EMEA with centralized vCPE enables the service provider to sell the same powerful tools they are using as hosted services to the enterprises, leading to enterprises becoming more efficient.
It is indeed very exciting times to work in the mix of network and software. Things happening are making networks more efficient and opening up for whole new business cases. Bringing me to ask you, what will you get out of NFV?
Topics: Expert Insights