Abstract: Methods and apparatuses for using a neural network based model to predict an output port for a destination Internet Protocol (IP) address in a network are described. Some embodiments can construct an untrained model comprising a graph neural network (GNN), a first artificial feed-forward neural network (ANN), and a second ANN. Next, the embodiments can train the untrained model to obtain a trained model by: training the first ANN using at least IP addresses of destination nodes in the network, training the GNN using at least an adjacency matrix of the network and initial node features computed using the IP addresses of destination nodes in the network, and training the second ANN by combining the output of the first ANN and an output of the GNN using an attention mechanism. The embodiments can then use the trained model to predict the output port for the destination IP address.

Abstract: Methods and apparatuses for using a neural network based model to predict an output port for a destination Internet Protocol (IP) address in a network are described. Some embodiments can construct an untrained model comprising a graph neural network (GNN), a first artificial feed-forward neural network (ANN), and a second ANN. Next, the embodiments can train the untrained model to obtain a trained model by: training the first ANN using at least IP addresses of destination nodes in the network, training the GNN using at least an adjacency matrix of the network and initial node features computed using the IP addresses of destination nodes in the network, and training the second ANN by combining the output of the first ANN and an output of the GNN using an attention mechanism. The embodiments can then use the trained model to predict the output port for the destination IP address.

Abstract: Methods and apparatuses for building a programmable dataplane are described. Specifically, the programmable dataplane can work on a list of identifiers, such as those part of OpenFlow 1.5. Specifically, the programmable dataplane can be built by creating a virtual network graph at a controller node using binary identifiers such that a node is broken into an n-ary tree and the tree has 1×2 or 1×1 nodes.

Abstract: Methods and apparatuses for building a programmable dataplane are described. Specifically, the programmable dataplane can work on a list of identifiers, such as those part of OpenFlow 1.5. Specifically, the programmable dataplane can be built by creating a virtual network graph at a controller node using binary identifiers such that a node is broken into an n-ary tree and the tree has 1×2 or 1×1 nodes.

Abstract: A network capable of being used in a datacenter is described. The network can comprise a set of optical fiber rings, wherein each optical fiber ring carries data traffic on multiple wavelengths, and wherein each optical fiber ring is partitioned into multiple sectors. In some embodiments, each sector in the multiple sectors can comprise: (1) only one add-wavelength-selective-switch (add-WSS) communicatively coupled to only one optical fiber ring in the set of optical fiber rings, wherein the only one add-WSS is used for sending all data traffic that originates from the sector and is destined to other sectors; (2) an add-electro-optical-switch (add-EOS) communicatively coupled to the add-WSS; (3) a set of drop-wavelength-selective-switches (drop-WSSs) communicatively coupled to the set of optical fiber rings, wherein the set of drop-WSSs are used for receiving data traffic from other sectors; and (4) a drop-electro-optical-switch (drop-EOS) communicatively coupled to a drop-WSS in the set of drop-WSSs.

Abstract: A network capable of being used in a datacenter is described. In some embodiments, the network can comprise a set of optical fiber rings, wherein each optical fiber ring carries data traffic on one or more wavelengths, and wherein each optical fiber ring is partitioned into multiple sectors. A reconfigurable optical add-drop multiplexer (ROADM) can be coupled to at least one optical fiber in each of at least two sectors. An electro-optical-switch can be coupled to each ROADM in each of the at least two sectors. A set of switches can be coupled to each electro-optical-switch in each of the at least two sectors. The set of switches can comprise a first layer of aggregation switches that is coupled to a second layer of edge switches, wherein the edge switches can be coupled to servers in a datacenter.

Abstract: A network capable of being used in a datacenter is described. The network can comprise a set of optical fiber rings, wherein each optical fiber ring carries data traffic on multiple wavelengths, and wherein each optical fiber ring is partitioned into multiple sectors. In some embodiments, each sector in the multiple sectors can comprise: (1) only one add-wavelength-selective-switch (add-WSS) communicatively coupled to only one optical fiber ring in the set of optical fiber rings, wherein the only one add-WSS is used for sending all data traffic that originates from the sector and is destined to other sectors; (2) an add-electro-optical-switch (add-EOS) communicatively coupled to the add-WSS; (3) a set of drop-wavelength-selective-switches (drop-WSSs) communicatively coupled to the set of optical fiber rings, wherein the set of drop-WSSs are used for receiving data traffic from other sectors; and (4) a drop-electro-optical-switch (drop-EOS) communicatively coupled to a drop-WSS in the set of drop-WSSs.

Abstract: Methods and apparatuses for merging continuity check messages (CCMs) are described. Some embodiments determine multiplexer and de-multiplexer nodes in a network for multiplexing and de-multiplexing CCM traffic. One embodiment creates an optimization problem which when solved identifies nodes in the network that should be configured as multiplexer nodes to multiplex multiple CCMs into a group CCM and/or as de-multiplexer nodes to de-multiplex a group CCM into multiple CCMs. This embodiment uses the solution of the optimization problem to configure nodes in the network as multiplexer nodes and/or as de-multiplexer nodes. Another embodiment determines weights for different paths in the network that can be used for merging CCM traffic, and then merges the CCM traffic based on these weights.

Abstract: A network capable of being used in a datacenter is described. In some embodiments, the network can comprise a set of optical fiber rings, wherein each optical fiber ring carries data traffic on one or more wavelengths, and wherein each optical fiber ring is partitioned into multiple sectors. A reconfigurable optical add-drop multiplexer (ROADM) can be coupled to at least one optical fiber in each of at least two sectors. An electro-optical-switch can be coupled to each ROADM in each of the at least two sectors. A set of switches can be coupled to each electro-optical-switch in each of the at least two sectors. The set of switches can comprise a first layer of aggregation switches that is coupled to a second layer of edge switches, wherein the edge switches can be coupled to servers in a datacenter.

Abstract: Systems and techniques for processing and forwarding packets are described. Specifically, some embodiments can include a receiving mechanism, a determining mechanism, a generating mechanism, and a sending mechanism. The receiving mechanism can be configured to receive a first packet that is to be routed from a first node in a first Autonomous System (AS) to a second node in a second AS. The determining mechanism can be configured to determine a set of bits that encodes a route in an n-ary tree that includes the first node and a root node in the first AS. The generating mechanism can be configured to generate, based on the first packet, a second packet that includes the set of bits and an identifier associated with the second AS. The sending mechanism can be configured to send the second packet.

Abstract: Methods and apparatuses for allocating backbone virtual local area network identifiers (BVIDs) to connection requests in a provider backbone bridging-traffic engineering (PBB-TE) network are described. During operation, a system (e.g., a network management system) can receive a set of connection requests for a set of connections that are desired to be set up in the PBB-TE network. Next, the system can allocate a BVID to each connection request in the set of connection requests so that, once the set of connection requests are set up in the PBB-TE network, conflicting connections in the PBB-TE network have different BVIDs.

Abstract: Methods and apparatuses for merging continuity check messages (CCMs) are described. Some embodiments determine multiplexer and de-multiplexer nodes in a network for multiplexing and de-multiplexing CCM traffic. One embodiment creates an optimization problem which when solved identifies nodes in the network that should be configured as multiplexer nodes to multiplex multiple CCMs into a group CCM and/or as de-multiplexer nodes to de-multiplex a group CCM into multiple CCMs. This embodiment uses the solution of the optimization problem to configure nodes in the network as multiplexer nodes and/or as de-multiplexer nodes. Another embodiment determines weights for different paths in the network that can be used for merging CCM traffic, and then merges the CCM traffic based on these weights.

Abstract: Methods and apparatuses for allocating backbone virtual local area network identifiers (BVIDs) to connection requests in a provider backbone bridging-traffic engineering (PBB-TE) network are described. During operation, a system (e.g., a network management system) can receive a set of connection requests for a set of connections that are desired to be set up in the PBB-TE network. Next, the system can allocate a BVID to each connection request in the set of connection requests so that, once the set of connection requests are set up in the PBB-TE network, conflicting connections in the PBB-TE network have different BVIDs.

Abstract: Systems and techniques for processing and forwarding packets are described. During operation, a system can receive a packet on an input port. Next, the system can identify a set of bits in the packet that represents a route from a source node to a destination node in an n-ary tree. The system can then determine an output port based on a subset of the set of bits. Next, the system can determine whether the output port is free. If the output port is not free, the system can use a contention resolution mechanism to store the packet in an on-chip memory or an off-chip memory based on space availability and the packet's priority. If the output port is free, the system can send the packet through the output port.

Abstract: Systems and techniques for processing and forwarding packets are described. Specifically, some embodiments can include a receiving mechanism, a determining mechanism, a generating mechanism, and a sending mechanism. The receiving mechanism can be configured to receive a first packet that is to be routed from a first node in a first Autonomous System (AS) to a second node in a second AS. The determining mechanism can be configured to determine a set of bits that encodes a route in an n-ary tree that includes the first node and a root node in the first AS. The generating mechanism can be configured to generate, based on the first packet, a second packet that includes the set of bits and an identifier associated with the second AS. The sending mechanism can be configured to send the second packet.

Abstract: A method for allocating the use of an optical light-trail includes calculating a bid at a first node in the light-trail with consideration for the criticality of the node's need to transmit particular traffic on the light-trail. The method also includes transmitting the calculated bid to one or more of the other nodes included in the light-trail. In addition, the method includes receiving an acknowledgment from one or more of the other nodes included in the light-trail, each acknowledgement indicating whether the transmitted bid is higher or lower than a bid calculated by the node from which the acknowledgment originates. The method also includes determining, based at least on the content of acknowledgements received, whether the transmitted bid is the highest bid.

Abstract: Providing protection for a network includes establishing a light-trail through a sequence of nodes of an optical network, where the sequence of nodes is coupled by a first fiber and by a second fiber. Traffic is communicated through a plurality of connections of the light-trail, where a connection is operable to communicate traffic from a source node of the sequence of nodes to one or more destination nodes of the sequence of nodes. A failure of the light-trail is detected. A protection light-trail corresponding to the light-trail is established. The traffic of the plurality of connections is communicated through the protection light-trail.

Abstract: In accordance with the teachings of the present invention, wavelength selective switch (WSS) design configurations for mesh light-trails are provided. In a particular embodiment, a node included in an optical network comprises a first light access unit operable to add or drop local traffic and a first WSS assembly communicatively coupled to the first light access unit and comprising at least four WSSs, at least two input ports, and at least four output ports, the four WSSs configured to direct any input traffic in any wavelength received at one of the input ports to the first light access unit and/or to any of the output ports.

Abstract: In accordance with the teachings of the present invention, wavelength selective switch (WSS) design configurations for mesh light-trails are provided. In a particular embodiment, a node included in an optical network comprises a first light access unit operable to add or drop local traffic and a first WSS assembly communicatively coupled to the first light access unit and comprising at least four WSSs, at least two input ports, and at least four output ports, the four WSSs configured to direct any input traffic in any wavelength received at one of the input ports to the first light access unit and/or to any of the output ports.

Abstract: A method for allocating the use of an optical light-trail includes calculating a bid at a first node in the light-trail with consideration for the criticality of the node's need to transmit particular traffic on the light-trail. The method also includes transmitting the calculated bid to one or more of the other nodes included in the light-trail. In addition, the method includes receiving an acknowledgment from one or more of the other nodes included in the light-trail, each acknowledgement indicating whether the transmitted bid is higher or lower than a bid calculated by the node from which the acknowledgment originates. The method also includes determining, based at least on the content of acknowledgements received, whether the transmitted bid is the highest bid.