Fast reroute with serial OBS activation

Figure 3 illustrates our proposed scheme. The scheme is based on the Fast Re-route scheme discussed in [7]. In this approach, the failed headend, upon sensing the failure, tries to setup an alternate path itself. In the path setup phase, the DCN headend constructs an OBS control packet containing information about timing and wavelength of the data burst to follow. When its upstream neighbor receives the control packet, it extracts information about the source of the following data burst and initiates a cross-connect operation in the OXC that it controls. It then updates the control packet with burst information for the next node and forwards it on. The ingress forwards the control packet further on downstream toward the egress through every node on the alternate path. Thus, the whole path is setup using the Just Enough Time (JET) burst switching technique. [8,9,10] suggest that such a signaling scheme which couples data bursts in a high speed OTN and control packets in a slower DCN is indeed practicable. Since the transmitting node (i.e., the node immediately upstream to the failed link) does not wait for an acknowledgement of path setup before forwarding traffic on the alternate path, the time to setup the path is significantly reduced. Moreover, the traffic stream can be resumed even before all the nodes on the path have received the control packet, as long as it can be guaranteed that the data burst always lags behind the control packet. The time required to complete each phase of the restoration process is as follows:

Figure 3: Serial OBS activation architecture.

$$t_{detect} = t_{LOL}$$ (12)

$$t_{FIS} = 0$$ (13)

Figure 4: Calculating $t_{setup}$ for serial OBS-based activation.

As shown in Figure 3, the alternate path for this scheme begins at the failed-headend, goes through the ingress and ends at the egress. Thus, as per the figure, if the set of nodes in the alternate path is:

$\mathcal{P} =$ { $node_2, node_3, \cdots, node_7$}

then, order to guarantee that the data burst always follows the control packet, the following constraint (illustrated by Figure 4) needs to be satisfied:

$\forall~n \in \mathcal{P}: d + O_{n} > t_{n}$

where $d$ is the time by which the data burst is required to lag behind the control packet, $O_{n}$ is the time taken by the burst to travel to node $n$ through the OTN, and $t_{n}$ is the time taken by the control packet to travel to node $n$ through the DCN. Hence:

$$t_{setup} = {\rm max}(\forall~n \in \mathcal{P} : t_{n} - O_{n})$$ (14)

$$t_{restore} = t_{LOL} + {\rm max}(\forall~n \in \mathcal{P} : t_{n} - O_{n}) + t_{switch}$$ (15)