Access nodes of a large-scale network are arranged into a number of groups. The groups are arranged into a number of bands. Each distributor of a pool of distributors interconnects each access node of a selected group to at least one channel from each group of a selected band. A discipline of allocating the selected group and the selected band to a distributor ensures that each access node has: a number, approximately equal to half the number of groups, of parallel single-hop paths to each other access node of a same group; a number, approximately equal to half the number of bands, of parallel single-hop paths to each access node of a different group within a same band; and one single-hop path to each other access node of a different access band. To eliminate the need for cross connectors, geographically-spread distributors are arranged into geographically-spread constellations of collocated distributors.

Systems and methods for performing impairment compensation in point-to-multi-point communication systems are described. In a data snapshot mode, a hub node can send instructions to each communication node connected to the hub node to send a data snapshot of data being received and processed by the communication nodes at a particular time. In a trench line mode, a hub node sends instructions to each communication node to send trench line data back to the hub node. The hub node uses the data snapshot or trench line data to determine how to tune filter coefficients in the hub node to perform impairment compensation and improve performance of the communication system.

The disclosed system implements a bandwidth-reconfigurable optical interconnect, which couples optical signals between N interconnect inputs and N interconnect outputs. The system includes an arrayed waveguide grating router (AWGR), which provides cyclic, single-wavelength, all-to-all routing between N AWGR inputs and N AWGR outputs. The system also includes a wavelength-insensitive switch, which provides all-wavelength, all-to-all connectivity between N wavelength-insensitive inputs and N wavelength-insensitive outputs. The system additionally includes a wavelength-selective input switch, which selectively directs up to L wavelengths from each of the N interconnect inputs into a corresponding input of the wavelength-insensitive switch, wherein unselected wavelengths from each of the N interconnect inputs pass into a corresponding AWGR input. Finally, the system includes a wavelength-selective output switch, which selectively directs up to L wavelengths from each of the N wavelength-insensitive outputs into a corresponding interconnect output, wherein each of the N AWGR outputs pass into a corresponding interconnect output.

A method and apparatus for controlling an optical switch. The switch includes a switching fabric and optical amplifiers for amplifying optical signals. A configuration for the switching fabric is generated and implemented. The configuration indicates a set of optical paths between switching fabric input ports and the output ports. Optical path losses through the switching fabric vary based on the configuration. An amplifier control signal for controlling gains of the optical amplifiers, is also provided. The configuration for the switching fabric is generated based on the gains of the optical amplifiers, the amplifier control signal is generated based on the configuration for the switching fabric, or both.

Methods, systems, and apparatus, including an apparatus for generating clusters of building blocks of compute nodes using an optical network. In one aspect, a method includes receiving data specifying requested compute nodes for a computing workload. The data specifies a target arrangement of the nodes. A subset of building blocks of a superpod is selected. A logical arrangement of the subset of compute nodes that matches the target arrangement is determined. A workload cluster of compute nodes that includes the subset of the building blocks is generated. For each dimension of the workload cluster, respective routing data for two or more OCS switches for the dimension is configured. One-to-many switches are configured such that a second compute node of each segment of compute nodes is connected to a same OCS switch as a corresponding first compute node of a corresponding segment to which the second compute node is connected.

A method and apparatus for controlling an optical switch. The switch includes a switching fabric and optical amplifiers for amplifying optical signals. A configuration for the switching fabric is generated and implemented. The configuration indicates a set of optical paths between switching fabric input ports and the output ports. Optical path losses through the switching fabric vary based on the configuration. An amplifier control signal for controlling gains of the optical amplifiers, is also provided. The configuration for the switching fabric is generated based on the gains of the optical amplifiers, the amplifier control signal is generated based on the configuration for the switching fabric, or both.

An optical switch fabric comprises two or more optical switch elements. The optical switch elements are configured in a topology. A switch control has a plurality of bias control signals. The switch control can address one or more of the optical switch elements and can apply one of the bias control signals to bias of the addressed optical switch element to establish a switch setting. The topology and switch settings determine how each of one of the inputs is connected to each of one of the outputs of the optical switch fabric. The switch settings are determined by a machine learning process which includes a model creation. The model can be made to adapt dynamically during optical switch fabric operation.

Constellations of distributors interconnect access nodes to form a vast contiguous network. The access nodes are generally geographically spread and the constellations are generally geographically spread, however the distributors within each constellation are collocated. The access nodes are arranged into access groups. The access nodes of each access group interconnect through selected constellations, with each access node having a wavelength-division-multiplexed (WDM) link to each of the selected constellations, to form a three-stage network. The three-stage networks corresponding to the access groups are mutually fused so that an access node of any three-stage network has multiple paths, each traversing one distributor, to each other access node of the same three-stage network and a path to each other access node of the entire network traversing one distributor. The distributors are preferable configured as fast optical switches. The network is structured to provide global coverage without the need for conventional cross-connectors.

A system and method for a tunable optical delay line. The tunable optical delay line comprises a coarse delay portion that provides a coarse delay amount, the coarse delay portion including a coarse delay selection element in conjunction with a coarse delay element, the coarse delay selection element incorporated on-chip into a photonic integrated circuit (IC) component, the coarse delay element being disposed off-chip of the photonic IC component and interconnected with the coarse delay selection element; and a fine delay element that provides a fine delay amount, the fine delay element interconnected in series with the coarse delay selection element, the optical delay line being tunable to a target delay amount by agglomerating the coarse and fine delay amounts.

An optical transmitter includes: a plurality of client ports configured to receive a client signal from an end user device; a plurality of line ports configured to generate a line signal in which the client signal is stored, and transmit the line signal to an optical receiver; a switch configured to connect the plurality of client ports with the plurality of line ports; and a label provider configured to provide the client signal with a label for identifying a transmission destination in the optical receiver.