How it works: network function virtualization

    119

    Network function virtualization is a growing trend for companies looking to test and experiment with their networks without taxing their existing IT resources.

    In this edition of “How it works,” Ely Compean, SVP with NetAmerica Alliance, shows us the basic structure and application of network function virtualization.

    Compean explains that NFV still requires hardware at the base level, but the layers that are added on top are what make it different from traditional network function testing.

    The first NFV layer, the hypervisor, is a virtual machine manager that allows multiple operating systems to share a single hardware processor.

    Multiple operating systems are then layered on top of that. The operating systems can be any type from Linux to Windows.

    The next layer is the application. The application can be any part of the network stack such as a load balancer, Web server, BSS, OSS or provisioning system.

    If extended, it could also be a VPN application, router application, or anything else that is part of the network, according to Compean.

    On top of the application layer lies the application-programming interface, which Compean says, “is where the beauty and migration is happening because we have API and this is where all the developers are focusing now because we know that the API is king.”

    An API is a software-to-software interface, which allows applications to talk to each other without user intervention.

    “Everything that used to be provided by an independent vendor, the integration was a nightmare … but now, for example, with open stack or any of the other open-source systems out there, all the communication is being handled through XML, API,” he explains.

    While many networks are still in private clouds, Campean sees the move to public clouds as a growing trend because it allows vendors to keep costs down while increasing scale.

    “When you move to the public cloud,” he said, “then you have vendors that are doing cost per CPU that are so cost-effective for people like me who are looking to scale and reduce cost and make the market go faster because we have to deliver products quickly.”

    For more on NFV, check out NFV/SDN Reality Check with Dan Meyer.