GMPLS
Initialism of generalized multi-protocol label switching.
GMPLS: label switching that spans every network layer
Generalized Multi-Protocol Label Switching (GMPLS) is a control plane architecture that extends MPLS (Multi-Protocol Label Switching) beyond packet networks to manage switched circuits, optical wavelengths, and time-division multiplexed (TDM) channels with the same signaling framework. Rather than forwarding packets by inspecting their headers at each hop, GMPLS uses fixed-length labels to steer traffic through data plane forwarding elements, but it also provisions and controls non-packet interfaces such as optical fiber ports and TDM slots across a unified control plane.
The core problem GMPLS solves is the operational complexity of managing heterogeneous transport networks. A carrier network might contain packet switches (routers), optical cross-connects (OXCs), TDM multiplexers, and wavelength-selective switches all performing different forwarding operations. Without GMPLS, each device class requires its own provisioning protocol, management interface, and monitoring system. GMPLS unifies this by using a single signaling protocol (RSVP-TE or CR-LDP) to establish labeled switched paths (LSPs) across any forwarding paradigm: IP interfaces, optical wavelengths, fiber links, or TDM timeslots.
The term 'generalized' refers to the extension of MPLS concepts to non-packet forwarding. Where classic MPLS labels are 20-bit values in packet headers, GMPLS labels can represent optical wavelengths (frequency slots), TDM timeslot identifiers, or physical fiber connections. The label format and forwarding mechanics change, but the routing and signaling logic remains consistent. This abstraction allows a single control plane to operate across optical, TDM, and packet layers, which is essential for carrier-grade networks where Layer 1 and Layer 2 resources must scale alongside packet demand.
Practical deployment and constraints
GMPLS is deployed primarily in carrier backbone networks and metropolitan optical networks where traffic engineering and fast restoration are critical. It enables recovery times under 50 milliseconds by pre-computing alternate LSPs and switching to them on link failure without waiting for higher-layer protocols to converge. However, GMPLS assumes that the network has a control plane channel separate from the data plane; optical networks typically maintain an out-of-band management network or use an embedded supervisory channel for signaling. Equipment must support GMPLS-capable control plane processors, which adds cost and complexity compared to static optical switching.
Common operational issues include control plane partitions, where signaling connectivity fails while data plane forwarding continues, leaving LSPs stranded; label space exhaustion on high-radix switches; and mismatch between advertised optical wavelength capabilities and actual transponder tuning ranges. Interoperability remains challenging because device vendors implement GMPLS subsets differently, particularly in optical-layer constraint encoding and graceful shutdown procedures. Standards exist (IETF RFCs 3471, 3473, and related documents) but real-world networks often require vendor-specific extensions.