The routing considerations for TVR

Introduction Time-Variant Routing (TVR) introduces a scenario of calculating a path, or sub-path within a network, taking into account the timing of message transmission or receipt as an integral part of the overall route computation. introduces typical TVR use cases and analyses the routing impacts. The routing computation is influenced by the parameters include links, adjacencies and costs, etc. defines the YANG model which specifies a recurring schedule for changing the attributes of resources used for TVR scenario. This document introduces three modes of routing computation for TVR scenario. Each node influenced by the TVR parameters can computes the routes correctly, predictably and precisely. For the convenience, the node with scheduled variation of link or cost, is called Variable Node. The VN may or may not have routing capabilities, such as routing advertisement or routing computation.

Each node gets the models directly Each node in the TVR scenario gets the YANG model which defined in directly from the controller, or by any other ways such as configuration. The YANG model needs to be translated to the items of database, such as route-LSA in OSPF, or LSP in IS-IS. All the nodes need to understand the YANG model and do the translating work, then the nodes can compute the predictable routing table. In this case, whether the VN has or has not routing capabilities does not have influence. This mode applies to all the use cases described in .

VN advertises the variability by typical advertisement In this case, the VN has the routing capabilities, it can advertise the routing variability to all the other nodes in TVR scenario. Before the predicted time, the VN advertises variabilities include link, adjacency or cost to other nodes by typical advertisement. In this case, except the VN, other nodes need not to get or translate the model from controller or by configuration. But the advanced time needs to be ensured, and there may be computation latency for variable node. This mode applies to the use case described in section 4 in .

Helper advertises the variability by typical advertisement In this case, some nodes which are called helpers, can do the typical advertisement instead of the VN. The helper usually is the adjacency routing node to the VN. Before the predicted time, the helper advertises the variability from the perspective of itself to the other nodes in the TVR scenario. For example, in the predictable time, the link of VN is down, the helper advertises the link or adjacency variability of itself. The helper node and other nodes can compute the routing table correctly. In this case, except the helper node, other nodes need not to get or translate the model from controller or by configuration. But the advanced time needs to be ensured, and there may be computation latency for variable node. This mode applies to the use case described in section 3 and section 5 in .

IANA Considerations This document has no IANA actions.

Security Considerations TBD.