Jump to: navigation, search

Difference between revisions of "Networking-vpp/CandidateFeatures"

 
(2 intermediate revisions by one other user not shown)
Line 1: Line 1:
 
== Feature roadmap ==
 
== Feature roadmap ==
  
Here is the list of candidate features for the release of networking-vpp to coincide with 17.01:
+
Here is the list of candidate features for the release of networking-vpp to coincide with 17.01 and for future releases.  The ones we aim to release on the 17.01 platform are highlighted.
  
 
'''Core features'''
 
'''Core features'''
 
* LISP-GPE overlay support: this would use the FD.io ONE project (https://wiki.fd.io/view/ONE) as a basis - an implementation of the overlay protocol using the networking-vpp + etcd infrastructure rather than ODL + Honeycomb.
 
* LISP-GPE overlay support: this would use the FD.io ONE project (https://wiki.fd.io/view/ONE) as a basis - an implementation of the overlay protocol using the networking-vpp + etcd infrastructure rather than ODL + Honeycomb.
* Routing full layer 3 support (including NAT, Floating IP, ...) based on VPP as opposed to the Linux router.  This would be a swap-in component for the Neutron L3 agent, and does not require that networking-vpp is the enabled mechanism driver.
+
* Routing full layer 3 support (including NAT, Floating IP, ...) based on VPP as opposed to the Linux router.  This would be a swap-in component for the Neutron L3 agent. [[https://wiki.openstack.org/wiki/Networking-vpp/L3_routing_support|L3 Details]]
 +
* '''Security groups'''
 
<br />
 
<br />
  
 
'''Robustness'''
 
'''Robustness'''
* Agent restart, VPP restart
+
* '''Agent restart, VPP restart'''
* Support agent restart without restarting VPP, and vice versa (as of 2016-12-08 if one shuts down we restart the other rather than incrementally fix its config)
+
**Support agent restart without restarting VPP, and vice versa (as of 2016-12-08 if one shuts down we restart the other rather than incrementally fix its config)
* Automated testing for restart scenarios
+
** Automated testing for restart scenarios
* Unit testing of resync code (resyncs can fail in many different locations, so automated system testing is guaranteed not to catch every possibility)
+
** Unit testing of resync code (resyncs can fail in many different locations, so automated system testing is guaranteed not to catch every possibility)
 
* Support and automated testing of driver restart
 
* Support and automated testing of driver restart
* Automated testing for restart scenarios
+
** Automated testing for restart scenarios
* Unit testing of resync code
+
** Unit testing of resync code
 
* Automated testing of arbitrary startup order of any component
 
* Automated testing of arbitrary startup order of any component
* Unit tests for code components
+
* '''Unit tests for code components'''
* Automated in-VM testing of VPP deployment as part of the OpenStack gate
+
* '''Automated in-VM testing of VPP deployment as part of the OpenStack gate'''
 
<br />
 
<br />
  
 
'''High Availability'''
 
'''High Availability'''
 
* etcd cluster management with or without etcd-proxy
 
* etcd cluster management with or without etcd-proxy
* Fault injection to the etcd cluster during networking-vpp tests
+
* '''Fault injection to the etcd cluster during networking-vpp tests'''
 
<br />
 
<br />
  

Latest revision as of 15:18, 22 January 2017

Feature roadmap

Here is the list of candidate features for the release of networking-vpp to coincide with 17.01 and for future releases. The ones we aim to release on the 17.01 platform are highlighted.

Core features

  • LISP-GPE overlay support: this would use the FD.io ONE project (https://wiki.fd.io/view/ONE) as a basis - an implementation of the overlay protocol using the networking-vpp + etcd infrastructure rather than ODL + Honeycomb.
  • Routing full layer 3 support (including NAT, Floating IP, ...) based on VPP as opposed to the Linux router. This would be a swap-in component for the Neutron L3 agent. [Details]
  • Security groups


Robustness

  • Agent restart, VPP restart
    • Support agent restart without restarting VPP, and vice versa (as of 2016-12-08 if one shuts down we restart the other rather than incrementally fix its config)
    • Automated testing for restart scenarios
    • Unit testing of resync code (resyncs can fail in many different locations, so automated system testing is guaranteed not to catch every possibility)
  • Support and automated testing of driver restart
    • Automated testing for restart scenarios
    • Unit testing of resync code
  • Automated testing of arbitrary startup order of any component
  • Unit tests for code components
  • Automated in-VM testing of VPP deployment as part of the OpenStack gate


High Availability

  • etcd cluster management with or without etcd-proxy
  • Fault injection to the etcd cluster during networking-vpp tests


Telemetry

  • Define a list of metrics to be exported from VPP forwarders to tenants


Security

  • RBAC for etcd
  • TLS support for control connections
  • JSON Web Tokens support