Storage aware congestion management
A method of congestion mitigation may include determining whether a host is sending a read command or a write command to an NVMe controller, and in response to a determination that the host is sending the read command, transmitting the read command via a first transmission control protocol (TCP) connection between the host and the NVMe controller. The method may further include in response to a determination that the host is sending the write command, transmitting the write command via a second TCP connection between the host and the NVMe controller.
1. A method of congestion mitigation, comprising:
determining whether a host is sending a read command or a write command to a non-volatile memory express (NVMe) controller;
in response to a determination that the host is sending the read command, transmitting the read command via a first transmission control protocol (TCP) connection between the host and the NVMe controller; and
in response to a determination that the host is sending the write command, transmitting the write command via a second TCP connection between the host and the NVMe controller.
2. The method of claim 1 , further comprising:
receiving an internet protocol (IP) header at the host in response to the write command;
determining if an IP header includes a congestion encountered (CE) bit within an explicit congestion notification (ECN), the CE bit indicating congestion in a transmit (Tx) direction from the host; and
throttling write inputs/outputs (I/Os) over the second TCP connection.
3. The method of claim 2 , further comprising
at a TCP layer of a network stack of the host, contracting a write TCP connection window size; and
continuing to send additional read commands at a first pre-instructed rate.
4. The method of claim 3 , further comprising:
determining if the congestion in a Tx direction from the host has subsided; and
in response to a determination that the congestion in the Tx direction from the host has subsided, restoring the write TCP connection window size to a pre-contracted state, processing additional write commands at a second pre-instructed rate, or combinations thereof.
5. The method of claim 1 , further comprising:
receiving an IP header at the host in response to the read command;
determining if the IP header includes a CE bit within an ECN, the CE bit indicating congestion in a receive (Rx) direction towards the host; and
throttling read I/Os over the first TCP connection.
6. The method of claim 5 , further comprising
at a TCP layer of a network stack of the host, sending instructions to the NVMe controller to contract a read TCP connection window size; and
continuing to send additional write commands at a first pre-instructed rate.
7. The method of claim 6 , further comprising:
determining if the congestion in the Rx direction toward the host has subsided; and
in response to a determination that the congestion in the Rx direction toward the host has subsided, restoring the read TCP connection window size to a pre-contracted state, processing additional read commands at a second pre-instructed rate, or combinations thereof.
8. A non-transitory computer-readable medium storing instructions that, when executed, causes a processor to perform operations, comprising:
determining whether a host is sending a read command or a write command to a non-volatile memory express (NVMe) controller;
in response to a determination that the host is sending the read command, transmitting the read command via a first transmission control protocol (TCP) connection between the host and the NVMe controller; and
in response to a determination that the host is sending the write command, transmitting the write command via a second TCP connection between the host and the NVMe controller.
9. The non-transitory computer-readable medium of claim 8 , the operations further comprising:
receiving an internet protocol (IP) header at the host in response to the write command;
determining if an IP header includes a congestion encountered (CE) bit within an explicit congestion notification (ECN), the CE bit indicating congestion in a transmit (Tx) direction from the host; and
throttling write inputs/outputs (I/Os) over the second TCP connection.
10. The non-transitory computer-readable medium of claim 9 , the operations further comprising:
at a TCP layer of a network stack of the host, contracting a write TCP connection window size; and
continuing to send additional read commands at a first pre-instructed rate.
11. The non-transitory computer-readable medium of claim 10 , the operations further comprising:
determining if the congestion in the Tx direction from the host has subsided; and
in response to a determination that the congestion in the Tx direction from the host has subsided, restoring the write TCP connection window size to a pre-contracted state, processing additional write commands at a second pre-instructed rate, or combinations thereof.
12. The non-transitory computer-readable medium of claim 8 , the operations further comprising:
receiving an IP header at the host in response to the read command;
determining if the IP header includes a CE bit within an ECN, the CE bit indicating congestion in a receive (Rx) direction towards the host; and
throttling read I/Os over the first TCP connection.
13. The non-transitory computer-readable medium of claim 12 , the operations further comprising:
at a TCP layer of a network stack of the host, sending instructions to the NVMe controller to contract a read TCP connection window size; and
continuing to send additional write commands at a first pre-instructed rate.
14. The non-transitory computer-readable medium of claim 13 , the operations further comprising:
determining if the congestion in the Rx direction toward the host has subsided; and
in response to a determination that the congestion in the Rx direction toward the host has subsided, restoring the read TCP connection window size to a pre-contracted state, processing additional read commands at a second pre-instructed rate, or combinations thereof.
15. A system comprising:
a processor; and
a non-transitory computer-readable media storing instructions that, when executed by the processor, causes the processor to perform operations comprising:
determining whether a host is sending a read command or a write command to a non-volatile memory express (NVMe) controller;
in response to a determination that the host is sending the read command, transmitting the read command via a first transmission control protocol (TCP) connection between the host and the NVMe controller;
in response to a determination that the host is sending the write command, transmitting the write command via a second TCP connection between the host and the NVMe controller;
receiving a first internet protocol (IP) header at the host in response to the write command;
determining if the first IP header includes a first congestion encountered (CE) bit within a first explicit congestion notification (ECN), the first CE bit indicating congestion in a transmit (Tx) direction from the host;
throttling write inputs/outputs (I/Os) over the second TCP connection;
receiving a second IP header at the host in response to the read command;
determining if the second IP header includes a second CE bit within a second ECN, the second CE bit indicating congestion in a receive (Rx) direction towards the host; and
throttling read I/Os over the first TCP connection.
16. The system of claim 15 , the operations further comprising:
at a TCP layer of a network stack of the host, contracting a write TCP connection window size;
continuing to send additional read commands at a first pre-instructed rate;
determining if the congestion in the Tx direction from the host has subsided; and
in response to a determination that the congestion in the Tx direction from the host has subsided, restoring the write TCP connection window size to a pre-contracted state, processing additional write commands at a second pre-instructed rate, or combinations thereof.
17. The system of claim 15 , the operations further comprising:
at a TCP layer of a network stack of the host, sending instructions to the NVMe controller to contract a read TCP connection window size;
continuing to send additional write commands at a first pre-instructed rate;
determining if the congestion in the Rx direction toward the host has subsided; and
in response to a determination that the congestion in the Rx direction toward the host has subsided, restoring the read TCP connection window size to a pre-contracted state, processing additional read commands at a second pre-instructed rate, or combinations thereof.
18. The system of claim 15 , wherein the read command and the write command are transmitted from the host to the NVMe controller via a fabric.
19. The system of claim 15 , further comprising:
a plurality of write queues, the plurality of write queues partitioning the write command from a number of additional write commands; and
a plurality of read queues, the plurality of read queues partitioning the read command from a number of additional read commands.
20. The system of claim 15 , the operations further comprising, at a block layer of a network stack, enqueuing the read command and the write command into separate QMaps.