Jump to: navigation, search

Project Proposals:Maple

Name

Maple

Repo Name

maple

Description

The objective of this project is to integrate into ODL the Maple programming model, which is a programming model that simplifies SDN programming for end users, by hiding the low level details of flow tables, which are a major source of SDN programming complexity. In particular, under the Maple programming model, an end user programs a function f that determines, using a high level algorithmic policy, how packets should be handled in a network. Maple then automatically translates the the high level policy into the low level flow tables of individual switches. In this particular project, we will implement the translation using a tracing runtime system, which collects user function f decision dependency to generate the flow tables of individual OpenFlow switches. See Maple SIGCOMM'13 for more technical details on the Maple programming model and its tracing run time technique.

One key use case of Maple in ODL is for implementing network services (e.g. service chaining, access control, route computations) in a way that allows them to be separately implemented and easily composed into overall services. Maple enables this by allowing services to be composed at a logical level, rather than at a device-specific, rule level and by providing a rule compiler and runtime that automatically generates device-specific configurations to reflect the overall composed service.

File:MapleInODL.pptx File:MapleInODL.pdf

Scope

We aim to implement Maple using MD-SAL.

Southbound and Northbound API Requirements for OpenDaylight

The tracing runtime design uses packets to generate rules to be installed on individual switches. Although one may use techniques such as warm up, offline packets to trigger the generation, in this project, we focus on the reactive model. Hence, the Maple runtime will need to receive missed packets (i.e., packets not matching any flow rules) at an ingress switch. Such packets will be sent from the switch to ODL Southbound OF 1.3 plugin. The Maple run time will need to receive such packets. Extension to handle non-OF switches will be future work.

After generating flow rules, the Maple runtime will then need to update the flow rules at individual switches. This project will use the ForwardingRulesManager (FRM) that manages the rules stored in the MD-SAL datastore to achieve this task. FRM will push the updates through the OF 1.3 Southbound plugin to individual switches.

Usage by Applications of OpenDaylight

In this initial version, we design the user application programming structure as follows. The Maple service is configured to execute the network according to a single, top-level, user-defined Java class implementing the MapleApp interface. In particular, the MapleApp interface includes the user function f, which will be invoked by the Maple runtime, when the decision of the function f on a packet is needed. The top-level MapleApp may invoke functions in other classes, including other MapleApp classes. The function f, as well as any other functions (transitively) invoked by f may access system and user states through ODL MD-SAL datastore. Each Maple application may expose APIs for others to read and update states defined by the application. The states of the Maple applications should be stored using MD-SAL datastore.

The Maple service will provide an RPC and model-driven configuration for specifying the top-level MapleApp. When the top-level user MapleApp is changed, Maple will refresh forwarding plane state to implement the new MapleApp's desired behavior.

Resources Committed (developers committed to working)

Initial Committers

Vendor Neutral

This project will develop new code. Code will be made available for review by ODP and Linux Foundation after it has been approved by contributing organizations.

Meets Board Policy (including IPR)