Microsoft Technology Licensing, LLC

Metaswitch Networks Limited

Matthew Ian Ronald Williams

Michael Jeffrey Evans

Andrew Munro Caldwell

Colin Tregenza Dancer

Jonathan Eric Hardwick

David Charles Hotham

Benjamin Thomas Wright

Richard Hewitt

Christopher Neil Swindle

David Reekie

Joseph Ellis Darracott Powell

Richard John Whitehouse

Murray William Rogers

Leland Douglas Benson

Colin Reynolds

Sophie Stratton

Martin Taylor

Calum Sutherland Loudon

Peter Louis White

Steven Victor Balls

James Robert Andrew

Abhishek Pandey

Daniel Thomas O'brien

The techniques disclosed herein improve existing systems by receiving a packet to process and determining an associated packet processing graph for the packet. In response to determining that the associated packet processing graph contains an action comprising a set of rules for modifying the packet processing graph, a table to modify in the packet processing graph is accessed based on a table identifier in the action. The accessed table is modified by inserting one or more table rows into the packet processing graph and for each inserted table row, inserting a set of template actions containing prepopulated fields. The modified table is used to process subsequent packets.

Self-modifying packet processing graph

In operation, packets traverse the packet processing data structure, and the network processing represented by each object in the data structure is applied to each packet. From time to time, the packet processing data structure may need to be updated. Embodiments of the present disclosure provide for lock-free updating of a packet processing data structure by means of epoch-based garbage collection. In embodiments, a particular past packet processing epoch is considered to be no longer referenced by any cores when the sequence numbers recorded in each said memory location are different to the sequence number of that particular past packet processing epoch. The deletion thread checks both whether a past epoch is being referenced by any packets and whether it is being reference by any cores. Thus memory is safely freed without having any impact on any packet processing which may be occurring in parallel to the deletion thread.

Packet processing

A service orchestrator receives a definition of a change to a network service to be deployed over a plurality of sites of a communications network. The definition comprises a configuration schema with a plurality of fields, which are partitioned into a plurality of different configuration group types. The service orchestrator receives an instance of the configuration group type. The service orchestrator receives conditional convergence logic specifying the received instance each site is to use conditional on a current state instance of the configuration group type used at the site. The service orchestrator sends instructions to each of the sites based on the conditional convergence logic, the instructions comprising the received instance. The instructions are configured to trigger a site convergence agent at each site to deploy the change to the network service according to the received instance of the configuration group type and the conditional convergence logic.

Deploying a change to a network service

Key data is processed using a hash table data structure. Control data storage locations have a write-availability indicated by a first write-availability state indicating that the location is available to be written to, and a second write-availability state indicating that the location is not available to be written to. A temporary label value is stored in a selected control data storage location to render the selected location and a corresponding key data storage location inactive for read-out. The write-availability of the control data storage location is changed to the second write-availability state in association with the storage of the processed label value in the selected location. A read-out label value is obtained and the temporary label value stored in the selected control data storage location is over-written with the obtained read-out label value, to render the selected control data storage location and corresponding key data storage location active for read-out.

Processing and storing data using a hash table data structure

The techniques disclosed herein enable systems to enhance network packet processing functionality through an accelerated packet classifier algorithm. Network packet classifiers serve to evaluate characteristics of network packets against various rules to determine how to process the network packet (e.g., routing, rewriting data). In contrast to existing systems, the present techniques pack a plurality of rules into a rule set which can then be applied to a network packet as a cohesive unit. By leveraging characteristics of binary arithmetic, the network packet classifier can evaluate the network packet against multiple rules in parallel. For example, the network packet classifier can concurrently compare values encoded by the network packet against a plurality of target ranges defined by the rules. This enables the network packet classifier to complete a full rule set evaluation within a single pass of the network packet thereby improving computing efficiency.

Accelerated network packet classifier

A method performed by a packet processor comprises retrieving a first prefix length for a first lookup stage and a shorter second prefix length for a second lookup stage. The packet processor can receive instructions to program an action entry. The length of the action entry prefix is compared to the first prefix and second prefix length. If it is determined that the length of the action entry prefix is shorter than the second prefix length, the second prefix length is selected as an expanded prefix length. If it is determined that the length of the action entry prefix is longer than the second prefix length and shorter than the first prefix length, then the first prefix length is selected as the expanded prefix length. Expanded action entries are generated with expanded prefixes of the expanded prefix length and stored at the lookup stage associated with the expanded prefix length.

Prefix length expansion in packet processors

A native restart controller restarts services in a controlled manner. The services run in a cloud computing environment comprising a plurality of computing devices executing a plurality of Kubernetes pods comprising one or more containers. An indication is received of a deployment defining a service to be implemented in the cloud computing environment, the deployment including a desired number of instances of the service. An API is instantiated that is operable to provide a first state indicative of whether restarts are to be controlled by the restart controller and a second state indicating that a restart is required. In response to receiving an indication that restarts for the cluster are to be controlled by the restart controller and that a restart is required, the instances are restarted in a sequence in accordance with restart criteria associated with the service.

Restart controller

The techniques disclosed herein improve existing systems by receiving, by a packet processor from a control node of the communications network, a request to program operations for processing data packets, the request received with opaque data comprising a state of the control node. The packet processor stores the operations and the opaque data in a data store of the packet processor. The packet processor receives from the control node a subsequent request to retrieve the opaque data. The packet processor sends to the control node the operations and the opaque data that were stored in the data store of the packet processor. The opaque data is not stored in the control node.

Stateless control planes

User traffic is processed in a virtualised network. First and second VNFs are initialised in the same network namespace as each other in user space in a host and have access to a shared memory region of the host. The first VNF processes user traffic and the second VNF provides a user plane service in relation to user traffic processed by the first VNF. The first VNF is used to establish a point-to-point, shared-memory interface between the first and second VNFs and is used to classify incoming user traffic. In response to the first VNF determining based on the classifying, that the incoming user traffic is to be subject to the user plane service, the first VNF is used to store the incoming user traffic in the shared memory region of the host to enable the second VNF to provide the user plane service in relation to the incoming trier traffic.

Processing user traffic in a virtualised network

A packet classifier and methods of processing data packets in a data communications network can include receiving a data packet comprising packet header data; performing rule-based classification of the received data packet to generate rule-based classification data; and processing the data packet in accordance with the rule-based classification data.

Packet classifier

In one aspect, a control plane Interworking Function (IWF-C) is used to establish a Packet Data Unit (PDU) session in a data communications network. A create session request is received, for example from a Mobile Management Entity (MME). A create session management context request is generated by the IWF-C and transmitted to a Session Management Function (SMF). A create session management context request response is received from the SMF. A create session request response is generated by the IWF-C and sent, for example to the MME. In other aspects, control plane policy interworking and user plane interworking is provided between components of 4G and 5G networks.

Growing community of inventors

London, United Kingdom

Matthew Ian Ronald Williams