The Network Processor (NPU) Advantage
Why NPU based OpenFlow Solutions
(Please view this short video first.)
NoviFlow offers switching products/solutions into this space that offer unparalleled performance and OpenFlow capacity. Key characteristics include:
• 240 Gbps of genuine OpenFlow 1.3/1.4 throughput
• Up to 1 million wildcard-match flow table entries in up 60 tables
• Up to 3 million exact-match flow table entries in up to 60 tables
• Up to 14,000 flow-modifications per second
NoviFlow switches are full-on OpenFlow-only switches, and do not implement hybrid switch features. The main reason for this is performance. Because of the functional limitations of the Application Specific Integrated Circuits (ASICs) they utilize, hybrid switches are either limited to basic OpenFlow 1.0 functionality, or are forced to implement OpenFlow 1.2 or 1.3 by running vSwitches on the switch’s host processor, or general purpose add-on processor cards, sacrificing throughput, flow table size, and being limited to only the OpenFlow 1.0 matching fields and actions supported by these same ASICs.
Unlike hybrid solutions, NoviFlow OpenFlow switches are implemented on high performance Network Processors via optimized software. This enables NoviFlow to implement almost the entire OpenFlow specification, including all matching fields and all actions. It also permits our products to fully support multiple tables per pipeline and to offer OpenFlow 1.3 compliant flow management with full L2-L7 packet inspection and processing. NoviFlow’s switches can expose these functions via the OpenFlow Experimenter interface, or (in the case of actions that cannot be supported by the OpenFlow standard) via an application programmable interface (API). This has the added benefit of making it possible for the NoviFlow solution to be customized to support network management applications and end-user applications beyond what is defined in the OpenFlow standard.
In addition to supporting the latest releases of OpenFlow, NoviFlow switches also offer advanced customization capabilities including support for the OpenFlow Experimenter interface and the ability for L2-L7 packet inspection and switching.
NoviFlow Inc.’s high performance switching products are ideal for use as intelligent edge devices, especially for initial SDN deployments into a network as an overlay solution. Not only is the high bandwidth of NoviFlow switches needed to meet the traffic processing demands of edge devices, but support for massive flow entry tables of up to 500,000 entries enables traffic filtering and management of very large numbers of flows, and makes practical flow forwarding logic using such fine-grained matching criteria as device IDs, individual labels and VLANs, etc. Typical applications include white/black lists for use in security, spam prevention, data flow segregation to support regulatory requirements, flow aggregation to support transparent VLANs, etc. Using an overlay approach will allow network operators to implement a gradual integration of OpenFlow-based solutions into existing networks without requiring major forklift upgrades and without causing any interruption of presently operating networks.
To summarize, NoviFlow’s products were designed with three strategic goals in mind:
• Performance: provide genuine high capacity OpenFlow throughput to enable overlay solutions on existing networks, support legacy networks and facilitating fast adoption cycles.
• Scale: support very large flow tables for network edge applications, offering customers the opportunity to simplify use of flow-based logic, scale solutions across networks at an unprecedented level of granularity.
• Flexibility: Offer a programmable interface to enable OpenFlow compliant packet handling services using the Experimenter interface, and non-OpenFlow customized packet handling routines using network processor resources.