DECIMOTERCERA. OCTAVA

OBS combines the benefits of circuit switching and packet switching [29]. Data packets are aggregated at the ingress of the source node forming larger data units called a burst. The burst is preceded by a control packet as a header. The control packet is sent via a reserved optical channel at the beginning to setup a circuit to the destination. The burst body is then sent through this circuit. These bursts are to be sent to the egress of the intermediate nodes in order to reach its destination. OBS is used to facilitate sending data

over all-optical networks, where it is infeasible to use packet switching due to the impracticality of processing each packet electronically since it is extremely difficult to manufacture optical buffers that are able to store or process packets in all-optical networks [30]. The following three signalling protocols are used to set up paths to the different destinations.

2.1.3.1 Just In Time (JIT) Signalling Protocol:

In JIT protocol, the source node sends a setup message to the switches along the path to the destination node to setup a connection. This setup message is followed by an offset to give time for setup message processing and switch configuration to reserve a bandwidth for the soon-coming burst. Bursts are transmitted in order of the setup messages arrival times, where the burst which is belonging to the first coming setup message is served first as shown in Figure 2.4 [31].

Figure 2.4 Just In Time timing event timing diagram [31]

Four signalling schemes are used here to allocate bandwidth to the burst as follows:

2.1.3.1.1 Explicit Setup and Explicit Release

In this scheme, the source node sends a setup message to the switches leading to the destination node in order to allocate a path for the forthcoming burst. After this path has been allocated, the transmission of the burst starts. When this burst ends, a release

message is sent through all the hops to the destination in order to set this path free for the next burst. Figure 2.5 shows the event timing diagram for this signalling scheme.

Figure 2.5 Explicit Setup and Explicit Release Event Timing Diagram [31]

2.1.3.1.2 Explicit Setup and Estimated Release:

The setup message here has information of the forthcoming burst size. This enables the network to release the path after sending the corresponding burst as shown in Figure 2.6. Therefore, there is no need to send a release message from the source as it is the case in the previous scheme.

2.1.3.1.3 Estimated Setup and Explicit Release:

Here, the setup message holds information about the estimated time of the burst’s arrival; therefore, the path will not be initiated as soon as the setup message arrives and processed, but instead it will be allocated at a predetermined time as shown in Fig. 2.7. After the path has been initiated, the optical burst is sent, and a release message is sent to the path switches when this burst ends.

Figure 2.7 Estimated Setup and Explicit Release Event Timing Diagram [31]

2.1.3.1.4 Estimated setup and Estimated Release

In this scheme, the header contains information about the estimated burst arrival time and burst length. Here, whenever a setup message reaches a switch, this switch will wait until the burst arrival time is due, then the switch configures itself to connect the burst to the appropriate output port. No release message is required here since the estimated time of burst end is obtained from the setup message.

Figure 2.8 Estimated Setup and Estimated Release Event Timing Diagram [31]

2.1.3.2 Just Enough Time (JET) Signalling Protocol

Unlike JIT protocol, in this scheme the incoming bursts are not necessarily allocated paths in order of their setup messages arrival times. In this signalling protocol, the source node sends a setup message to the ingress switch; the switch in turn runs a void filling algorithm waiting for any other setup message coming during the offset interval of the first setup message. If a new setup message arrives during this interval, the switch checks the burst’s arrival time and length. If this burst fits in the time slice between its arrival time and a few moments before the first bit of the first burst start transmission, just giving time for the switch to reconfigure itself, then this burst is accepted and a path is allocated for it, otherwise it is refused and dropped. This protocol is the best known signalling protocol because it improves the network resources utilisation [32].

In Figure 2.9 , if a setup message of burst A arrives to the ingress switch at time
 and the setup message of burst B arrives at
, the switch compares the arrival times of

the two bursts stored in their setup messages. Supposing that burst B arrives and ends before the estimated arriving of burst A, the path for passing Burst B is to be setup before the path for burst A, even though the setup message of Burst B came later.

Figure 2.9 Just Enough Time event timing diagram [31]

2.1.3.3 Horizon Signalling Protocol

Here, each outgoing path (or wavelength) will be assigned a horizon time. The next burst will not be scheduled until the horizon time has been exceeded. As soon as the setup message is received, the horizon will be updated. Figure 2.10 supposes that the setup message of burst i was received at t1, the end time for burst i was estimated from the

setup message to be t4. Giving an extra time unit for the switch to be reconfigured, the

horizon will be t5. Then burst2 will not be scheduled for transmission unless its arrival

time is later than t5. Since burst2 starts transmission at t6, then it will be accepted. After

scheduling of the burst2 the horizon of this channel is updated to t8 to allow the

transmission of burst2 to be completed and the switch is prepared for the next transmission, supposing that switch configuration process consumes one time unit [33, 34].