In a Vector Packet Processing (VPP) environment, such as may be implemented for a mobile network Control and User Plane Separation (CUPS) mobile gateways, each VPP worker executes the same code, typically operating as graph nodes. Some graph nodes, such as session queue nodes (typically used for Lawful Intercept (LI) Transmission Control Protocol (TCP) processing), Encapsulating Security Payload (ESP) encryption nodes (typically used for Internet Protocol Security (IPSec) processing), Transport Layer Security (TLS) related graph nodes, or the like, can involve Central Processing Unit (CPU) intensive operations. For example, if LI or IPsec is enabled, all workers will have these graph nodes, thus, consuming CPU resources for all such workers, which can cause a significant performance impact. This proposal provides techniques through which all VPP workers may not run/include all graph nodes when used in mobile gateways. In particular, each of multiple CPUs/threads for VPP workers can be partitioned into specific work item categories (e.g., IPSec encryption, TCP optimization, etc.) that may be utilized to perform specific, per-work item operations in addition to user packet processing operations. Such techniques may provide the ability to improve performance for specific work items in a VPP mobile network environment.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Krishnamurthy, Harish Kumar, "DYNAMIC GRAPH NODE ALLOCATION FOR VECTOR PACKET PROCESSING WORKERS OPERATING IN A MOBILE NETWORK USER PLANE", Technical Disclosure Commons, (April 27, 2022)