What Is Optical Burst Switching?

Optical burst switching (OBS) is a network switching paradigm in which client packets are aggregated at the edge of the network into larger units called bursts, which then traverse the optical core without conversion to the electrical domain. OBS occupies a position between optical circuit switching and optical packet switching: it does not require the long-lived, quasi-static connections of circuit switching, nor does it demand the all-optical buffering and ultra-fast switching that optical packet switching requires. The concept was developed in the late 1990s as a candidate transport architecture for a wavelength-division multiplexed optical Internet, combining the bandwidth efficiency of packet networks with the all-optical transparency of circuit networks.

The defining characteristic of OBS is the separation of control information from the data payload. A burst header packet (BHP) is transmitted ahead of the data burst on a dedicated control channel, processed electronically at each intermediate node to reserve the needed wavelength and time slot. After a fixed offset time, the data burst follows on the data channel, passing through optical cross-connects without electronic termination.

Burst Assembly and the Control Plane

At an OBS ingress node, arriving client packets are sorted by destination and assembled into bursts according to assembly policies governed by a timer or a size threshold. Short timers reduce latency but produce smaller, less efficient bursts; larger burst sizes improve channel utilization at the cost of increased end-to-end delay. The assembled burst is preceded by its BHP, which carries the burst length, destination address, and wavelength assignment. The ingress node sends the BHP on the control channel, and after the offset interval, releases the burst onto the data channel. The ACM Digital Library paper introducing OBS as a paradigm for the optical Internet provides the foundational analysis of these assembly and scheduling mechanisms.

Reservation Protocols

OBS reservation protocols differ primarily in whether they use one-way or two-way resource reservation. In one-way, or tell-and-go, protocols such as just-in-time (JIT) and just-enough-time (JET), the BHP reserves resources at each hop as it propagates toward the destination, and the burst follows without waiting for an acknowledgment. JET minimizes the reserved time window by computing the exact offset needed to align the burst arrival with the end of the setup time at each node. Two-way protocols require a round-trip acknowledgment before the burst is transmitted, guaranteeing resource availability but introducing a latency penalty equal to the round-trip propagation time. One-way protocols dominate in practice because the additional latency of two-way reservation offsets the benefits of guaranteed delivery for most traffic classes. A comparative analysis of OBS architectures, including node and protocol trade-offs, appears in a ScienceDirect survey of optical burst-switched network architectures.

Node Architecture and Contention

An OBS core node consists of optical amplifiers, a wavelength demultiplexer, an optical switching fabric, a wavelength multiplexer, and an electronic control unit. The electronic unit processes BHPs and configures the switch fabric to route each incoming burst to the designated output wavelength and port before the burst itself arrives, using the offset time as setup headroom. Contention occurs when two bursts require the same output wavelength at the same time; common resolution strategies include deflection routing to an alternate path, wavelength conversion to a free channel, and fiber delay line buffering to defer one burst. The Springer Optical Burst Switching chapter surveys contention resolution methods and their impact on burst loss probability in mesh OBS networks.

Applications

Optical burst switching has applications in a wide range of network scenarios, including:

• High-capacity WDM core networks, carrying aggregated IP traffic all-optically
• Metro optical rings, handling bursty data flows between edge nodes
• Data center interconnects, providing high-bandwidth low-latency inter-rack connectivity
• Carrier-grade optical Internet backbones, as an alternative to MPLS-over-WDM
• Research testbeds for evaluating dynamic wavelength provisioning strategies