Hello all,

This email is to explain how the topic presented in this email [0] has been implemented in this patch [1].

Previously there was a symmetric boolean flag in the SFC, SFP, and RSP RPC data models. Only the SFP symmetric flag was being used to create symmetric service chains. In the patch I deprecated the symmetric flag in the SFC and RSP RPC data models, and left the SFP symmetric flag as it was before. Additionally, I added changes to start using the Service Function SF-type bidirectional flag (I changed the name from bidirectionality to bidirectional).

In this email thread [0] I originally mentioned we would use the SF-type symmetry flag, but based on discussions, we decided it was better to instead use the SF-type bidirectional flag.

Now there are 2 different ways as follows to make an RSP (Rendered Service Path) symmetric. A symmetric RSP means that when an RSP is created, an additional symmetric, reverse RSP will also be created since RSPs are directional. This is basically for the case when you want an upstream and a matching downstream RSP.

 * If any Service Function is of a SF-type that has the bidirectional
   field set true, then the resulting RSP will be symmetric
 * The SFP symmetric flag, if present, will /*override*/ the SF-type
   bidirectional fields.
     o If the SFP symmetric flag is present and is true, then a
       symmetric RSP will be created independently of how the SF-type
       fields are set.
     o If the SFP symmetric flag is present and is false, then a
       symmetric RSP will not be created.


The motivation behind using the SF-type bidirectional flag is to allow for flexibility when creating RSPs. With this change, an operator doesnt need to know details about the SFs in the service chain. The SF providers set the SF-type configuration accordingly, and internally SFC will automatically determine if the resulting RSP needs to be symmetric or not.

Regards,

Brady

[0] https://lists.opendaylight.org/pipermail/sfc-dev/2016-November/003599.html
[1] https://git.opendaylight.org/gerrit/#/c/49546/

_______________________________________________
sfc-dev mailing list
[email protected]
https://lists.opendaylight.org/mailman/listinfo/sfc-dev

Reply via email to