Survivable hybrid optical/electrical data center networks using loss of light detection

1. A system for reliable data communication, comprising:

an electrical switching network with packet granularity of switching and store-and-forward forwarding services to network traffic;

an optical switching network providing circuit granularity of switching and optical bypassing services to network traffic,

wherein the electrical and optical switching networks comprise distributed and multi-hop data communication; and

a loss-of-light detection module coupled to the optical switching network to detect an optical failure point at each optical switching hop;

a network controller coupled to the electrical switching network, the optical switching network, and loss-of-light detection module, the network controller receiving failure data using network control protocol messages and reroutes traffic to avoid the optical failure point; and

a finite state machine (FSM) to detect failure,

wherein in case of a failure, after power up, the FSM places the optical switching network components in a self-check state and checks all port functionality and if the self-check fails, the optical switching network components send a signal to the network controller, stay in the self-check state and wait for repair or replacement, and if the self-check succeeds, the optical switching network components enter an active state, and an optical port connection map is configured and changed by the network controller in real time according to a traffic matrix or traffic matrix estimation, and if loss-of-light event is detected, the optical switching network components enter fault state,

wherein the network controller locates and isolates the failure point in the network, and triggers an alarm for recovering from the failure within a predetermined period.

2. The system of claim 1 , wherein the network comprises server racks, top of rack (TOR) switches, aggregation switches, and core switches.

3. The system of claim 1 , comprising switches in a FatTree based DCN with an electrical switching fabric and an optical switching fabric connected inside the switches in any layer.

4. The system of claim 1 , wherein optical power at each input port is tapped off by optical splitters for a predetermined amount sufficient to trigger a photodetector (PD) at each input port, wherein the PD and a PD-array controller are responsible of identifying the loss-of-light triggering condition and report failures.

5. The system of claim 1 , wherein optical power is tapped off at output ports of the switch rather than input ports.

6. The system of claim 1 , wherein the switch passes a fault-test on a corresponding port pair and transfers to a recover state as a loss-of-light signal is caused by a link failure from upstream optical fiber links.

7. The system of claim 1 , wherein both input ports and output ports are equipped with optical splitters or power tapping components to detect whether the loss of light detection comes from an internal node failure or an external link failure.

8. A method for data communication, comprising:

forming a hybrid that includes an electrical switching network with packet granularity of switching and store-and-forward forwarding services to network traffic and an optical switching network providing circuit granularity of switching and optical bypassing services to network traffic, wherein the electrical and optical switching networks comprise distributed and multi-hop data communication; and

positioning loss-of-light detection modules in optical switches of the optical network to detect an optical failure point at each optical switching hop;

applying local control of optical switching modules in the optical switching network with a network controller coupled to the electrical switching network, the optical switching network, and a loss-of-light detection module, wherein the network controller receiving failure data using network control protocol message;

detecting loss-of-light in a failure point in the optical switching network and rerouting traffic to avoid the optical failure point; and

handling a link or node failure point in the optical switching network with one or more finite state machines,

wherein in case of a failure, after power up, placing optical switching network components in a self-check state and checking all port functionality and if the self-check fails, the optical switching network components send a signal to the network controller, stay in the self-check state and wait for repair or replacement, and if the self-check succeeds, the optical switching network components enter an active state, and an optical port connection map is configured and changed by the network controller in real time according to a traffic matrix or traffic matrix estimation, and if loss-of-light event is detected, the optical switching network components enter fault state,

wherein the network controller locates and isolates the failure point in the network, and triggers an alarm for recovering from the failure within a predetermined period.

9. The method of claim 8 , comprising aggregating server traffic with electrical switches and entering an optical switching domain providing optical bypassing.

10. The method of claim 8 , comprising receiving multiple loss-of-light signals and determining a switch with a first fault, and notifying downstream switches of the first fault switch.

11. The method of claim 10 , wherein the first fault switch is a far-most switch to a light path end.

12. The method of claim 10 , wherein the first fault switch handle the fault while other switches reset corresponding ports in a previous light path to prepare for a new connection.

13. The method of claim 8 , wherein a switch passes a fault-test on a corresponding port pair and transfers to a recover state as a loss-of-light signal is caused by a link failure from upstream optical fiber links.