By: Sasa Nijemcevic
To support the cloud era, network service fulfillment must evolve to become on-demand, agile and able to rapidly scale operations. This requires moving to a carrier SDN-controlled environment that simplifies network complexity by abstracting it and enabling automated provisioning to IP/MPLS, carrier Ethernet and optical network services delivery. Moreover, to make on-demand operations profitable, carrier SDN needs to identify the best use of available network resources and meet service requirements, including service assurance, through network-aware dynamic service provisioning.
The cloud makes life so much easier for enterprises and consumers. Scale your business as demand grows. Choose the content, services and products you want, whenever you want and experience near-instant gratification. And yet, ironically, while making everyone else’s life easier, the cloud puts extraordinary demands on the network and, by extension, network operators. All this scaling up and down at web speed, especially in a global marketplace, creates enormous and unpredictable fluctuations in network traffic.
The telco operating environment wasn’t traditionally architected to be this agile — anything but. In fact, the major innovations in networking required to meet the challenges of the cloud era started in the data center LAN. The tremendous fluctuations that occurred as servers were spun up and down with globally trending demands, especially in web services, also had to be matched by the local area network capacity in the data center. The two principle innovations that were made in order to adapt to the demands of cloud services were virtualization and software-defined networks (SDN).
In essence, instead of running on dedicated network hardware, the internal data center network or LAN was abstracted as a software-layer and hosted on commercial off-the-shelf servers (COTS) as virtual machines (VMs), usually running Linux and VMware or the equivalent. These were, of course, the same virtualized servers on which the data center was running all its operations, so this wasn’t a big leap from an operational perspective. If a global-level spike in demand required spinning up hundreds of virtual machines as web servers, the network that supported all this traffic could be spun up too, on a different set of VMs. Routers are, after all, purpose-built network computers and can be virtualized.
This was several orders of magnitude simpler in a data center LAN, but the principle is identical when applied to the carrier WAN. To achieve this kind of agility and scalability, SDN can and is being applied to the wider network as “carrier” SDN. Similarly, virtualization is also being applied to network hardware as network function virtualization (NFV) — adding excess, rapidly scalable capacity to the existing networks using software-based abstractions of the network’s central functions, again running on COTS.
The important difference from the purpose-built data center LAN is that NFV in the WAN will, initially at least, mostly provide additional capacity for rapid scaling. The legacy optical, switching and routing network infrastructure will still be with us for a long-time. Thus it is SDN more that NFV that is having the biggest effect in the telco world. Carrier SDN can abstract and orchestrate all the various network functions, whether they are “virtual” (being hosted on COTS-based VMs) or “physical”, as with dedicated network hardware. Carrier SDN has to be able to apply across both virtual and physical network resources.
This is an enormous undertaking for network operators precisely because of the complexity of the WAN infrastructure. It is very much a work in progress and will take years to complete. It would in many ways simplify the task to start all over, but the size of investment in the current infrastructure prohibits it. Thus, the industry is in the midst of a massive effort to abstract the network, mostly as it currently exists, capturing in a software layer all of the virtual and physical network objects, flows and connections so that it is possible to fulfill network-based services in a matter of minutes and hours, not days and weeks.
While implementing carrier SDN is the first step in achieving the agility and scalability to support on-demand cloud-based services, another step needs to be taken quite quickly. The problem, operationally, is that having abstracted the virtual and physical network resources as objects, flows and connections, it is now the operator’s IT team that has to deal with the very same complexity that slowed down the network operations teams previously. Some speed has been gained because they are setting up the services in software, but some of that same speed has been lost because the IT teams don’t fully understand the complexity of the underlying network elements.