| commit | 0db4c81889d8231e4b89832378815e0e7a5f2291 | [log] [tgz] |
|---|---|---|
| author | khenaidoo <knursimu@ciena.com> | Wed May 27 15:27:30 2020 -0400 |
| committer | khenaidoo <knursimu@ciena.com> | Fri May 29 22:37:01 2020 -0400 |
| tree | cea688df8fa14faa65fa4dedfb1d249226d6f426 | |
| parent | 3b3f4a668b05fb6d938b4d3f8cfa196454d7e7fa [diff] |
[VOL-2995] Improve Core performance This commit consists of the following changes with the aim to improve the Core performance: 1) Use a hybrid approach of pre-route calculation and route calculation on-demand. For example, attempts to pre-calculate routes will be done whenever a nni/uni port is discovered. The attempt may fail if there are not enough ports to generate a route. When a flow is received and the route is not available then only the route relevant to that flow will be created on demand. 2) Changes some of the route calculation flow such that the process does not need to go and grab the latest version of the device which could lead to higher latency, expecially if that device is busy with other processing. 3) Change the logic when reporting added ports to ONOS such that routes are calculated (at least an attempt made) before sending a port create notification to ONOS. 4) Move peer port creation into its own go routine thereby removing the lock on a child device much earlier. 5) Wait until a request for port capability is received before removing the lock on a device. A better approach is required where the adapter will need to report the port capability along with the port creation event. However, this require another Jira as changes will be required in the API. 6) Remove some unnecessary proto.clones. Those are the obvious ones. Removal of other proto.clones will be done in a separate commit. 7) Fix a core panic when concurrent requests are made to the route map Change-Id: I2bafc99dbf10d7026572a44af0b88a31b5eb1887
Voltha aims to provide a layer of abstraction on top of legacy and next generation access network equipment for the purpose of control and management. Its initial focus is on PON (GPON, EPON, NG PON 2), but it aims to go beyond to eventually cover other access technologies (xDSL, Docsis, G.FAST, dedicated Ethernet, fixed wireless).
Key concepts of Voltha:
Control and management in the access network space is a mess. Each access technology brings its own bag of protocols, and on top of that vendors have their own interpretation/extension of the same standards. Compounding the problem is that these vendor- and technology specific differences ooze way up into the centralized OSS systems of the service provider, creating a lot of inefficiencies.
Ideally, all vendor equipment for the same access technology should provide an identical interface for control and management. Moreover, there shall be much higher synergies across technologies. While we wait for vendors to unite, Voltha provides an increment to that direction, by confining the differences to the locality of access and hiding them from the upper layers of the OSS stack.
You can start by reading the published documentation.
Another great way is to check out the BUILD.md file to see how you can build it, run it, test it, etc.
Contributions, small and large, are welcome. Minor contributions and bug fixes are always welcome in form of pull requests. For larger work, the best is to check in with the existing developers to see where help is most needed and to make sure your solution is compatible with the general philosophy of Voltha. Please check out the contributing page on the documentation.