The promise of the carrier cloud is more than just the advantages of mobility and scale. It also is about automation and flexibility. With the cloud, computing applications can scale up and down as needed, and in an automated fashion.
This fundamentally changes how networks perform.
The problem is that much of this advantage is lost due to networks not having this flexibility. While the carrier cloud can scale and adjust applications and servers on the fly, carrier Ethernet cannot keep pace because setting up the network connections can take days or even weeks.
Freeing up the bottleneck requires software-defined networking (SDN), which takes virtualization to the network level and creates the same flexibility and automation for networking as currently exists with software applications.
In an SDN data center network, connections are initiated by virtual machines in seconds and then dismantled automatically when no longer needed.
But as the cloud expands to encompass multiple data centers and users increasingly concentrated in metro and wide area networks, noted a recent Alcatel-Lucent (News - Alert) white paper, The Cloud-Optimized MAN and WAN, applications will expect the carrier network connections that underpin this new, distributed cloud fabric to leverage SDN and be just as easy to set up and consume. In short, as the paper details, metro networks need to be cloud-optimized.
Getting there requires augmenting existing traffic engineering processes “with a dynamic resource management capability that can ensure network efficiency, resiliency and availability in the face of rapidly changing cloud connectivity needs,” according to the paper.
Preparing the network for this SDN model requires a multi-layer SDN framework that partitions carrier networks into two major components: network virtualization and automation, and programmable IP/optical transport.
The programmable IP/optical transport layer is primarily about delivering highly reliable, high performance packet/optical, Ethernet, IP and VPN transport. According to Alcatel-Lucent, it needs to include:
- Specialized NPUs and optical hardware deliver Layers 0 to 3 grooming, switching and forwarding capabilities at an optimal price for performance
- Proven, distributed protocols that are embedded within network hardware to ensure network scale, stability and resiliency.
- Open APIs such as OpenFlow to allow standards-based applications and SDN controllers to monitor and control network resources.
The Network virtualization and automation layer provides a high-level, abstracted view of the network to applications, and has real-time visibility and control of all network resources.
Alcatel-Lucent believes that it needs SDN controllers that are used to provision all elements in one or more network layers, policy-based service provisioning that allows network tasks and services to be defined as policies so that they can be instantiated faster and on a mass scale, policies that can be used to combine many lower-level network tasks into a higher level function that shields applications from the unnecessary complexity of vendor-specific, low-level provisioning, and resources for discovery and control.
If the metro networks are to be cloud-optimized, as Manish Gulyani, Vice President Product Marketing, Alcatel-Lucent highlights in a recent corporate blog, operators need to evaluating and ultimately implementing how to adjust their networks and do it quickly. As Gulyani says, they need to shift metro networks into high gear.
Edited by Peter Bernstein