Self-Serve
Bandwidth
On Demand
For Internet Exchanges (IXPs) and ISPs, self-provisioning bandwidth under SDN control is a new bandwidth sales model that lets operators or users set up bandwidth in minutes rather than months.
Bandwidth on demand drivers and challenges
Network operators face increasing pressure from their customers for faster delivery times and reduced costs for the services being provided. Customers’ expectations have shifted for network resources—they believe that resources should only be made available when they are needed, and only charged for when they are actually used.
The IXPs and ISPs that deliver a bandwidth-on-demand model will win in the eyes of the customers. But, traditional networks are too complex and not dynamic enough to deliver network services under this new paradigm.
SDN is being used to solve the problem. It is critical for the SDN data plane to have the features and performance suited to this kind of WAN application.
SDN data planes power policy-driven management of bandwidth
Network operators need to meet higher user demand for flexibility and performance, even as increasing traffic volume consumes network resources. This requires the ability to dynamically prioritize packets for immediate, real-time bandwidth and congestion management using open, standard APIs or customer designed control software.
Corsa’s SDN data planes allow network operators and their users to gain direct control over bandwidth provisioning for their applications at a very granular, per-flow level. At the same time, they experience greatly improved network performance with reduced data transfer times and congestion.
Features suited for bandwidth on demand
Corsa’s SDN data planes allow network operators to deliver a complete set of software defined Layer 2 packet services without compromising the performance or reliability of their networks. Corsa devices provide a very high level of performance in a highly programmable and flexible open SDN data plane and feature:
- Complete control of bandwidth and congestion handling mechanisms via open standard OpenFlow 1.3+ APIs, including OpenDaylight, ONOS, RYU, and many others
- Highly granular classification in multiple match-action tables based on Layer2/3/4 fields to match flows
- RFC4115 Two-Rate, Three-Color Meter table that define bandwidth profiles for each flow
- Full line rate forwarding on all 10G, 40G and 100G interfaces, deep interface buffers, large OpenFlow classification tables and high flow modification rates
- Full SDN QoS classification, metering, queuing, and traffic handling capabilities
- User-configurable rates and weights to make packet processing flexible and powerful
By providing policy-aware, software driven provisioning, IXPs and ISPs are able to provide completely self-help, billable and provisionable bandwidth allocation and congestion management in their network.
Open programmable interfaces
Of the list above, the open programming interfaces deserve further explanation. Different customers require open programming interfaces for the network at different layers, depending on their requirements and comfort level. Open programmability is at the core of Corsa devices, and why we enable programmable interfaces at all layers—from the hardware layer to the network application layer. Network operators have the choice to use open source software or Corsa supplied software. By providing all the additional functionality for higher abstraction layers, RESTful APIs for all levels of configuration as well as built in network protocol stacks for most common services, Corsa significantly reduces the cost of software development that network operators have to incur to either adapt existing open source software to their needs, or to write complete software systems from scratch for network applications where open-source software does not exist. Most importantly, network operators avoid vendor lock-in.
SDN network architects can easily control network bandwidth allocation for each flow through their SDN controller’s southbound OpenFlow interface, and signal changes directly to the hardware forwarding the traffic. In turn, this type of network control can be presented via northbound APIs to provisioning, orchestration or other policy systems to enable on-demand bandwidth reservation or near real-time QoS policy changes. Potentially these APIs can even be exposed to the user applications directly in the form of new services, such as application admission control or dynamic bandwidth reservation. This level of programmatic control leads not only to reduced transfer times for large flows, but also gives users visibility and predictability for the performance of their applications.
Learning Center
