Selective network traffic throttling
Exemplary methods, apparatuses, and systems a management server receiving, from a switch, an indication of network congestion at the switch. The indication includes one or more contributors to the network congestion. In response to the indication of network congestion, the management server identifies one or more hosts running the identified contributor(s) and transmits an instruction to the host(s) to restrict network traffic originated by the contributor(s). For example, the restriction may include a reduction in the contributor's I/O queue depth. If the reduction in traffic does not remedy the congestion, the management server may receive a further indication of congestion and further instruct the host(s) to restrict identified contributor network traffic. For example, the contributors may have their I/O queue depth reduced exponentially until congestion is relieved. Upon receiving an indication that the congestion is relieved, the management server instructs the host(s) to remove the restriction on network traffic.
1. A computer-implemented method, comprising:
receiving, from a switch, an indication of network congestion at the switch, wherein the indication includes a contributor to the network congestion;
identifying, in response to the indication of network congestion, a host running the contributor; and
transmitting, to the host in response to the indication of network congestion, a first instruction to restrict network traffic originated by the contributor.
2. The computer-implemented method of claim 1 , wherein the contributor is an application running on the host.
3. The computer-implemented method of claim 1 , wherein the contributor is a data compute node running on the host.
4. The computer-implemented method of claim 1 , further comprising:
receiving, from the switch, an indication of network congestion relief; and
transmitting, to the host in response to the indication of network congestion relief, a second instruction to remove the restriction on network traffic originated by the contributor.
5. The computer-implemented method of claim 1 , wherein the restriction is a reduction in input/output (I/O) queue depth.
6. The computer-implemented method of claim 1 , wherein the contributor to the network congestion is a largest contributor to the network congestion.
7. The computer-implemented method of claim 1 , wherein the contributor to the network congestion is determined to contribute above a threshold amount of traffic to the network congestion.
8. A non-transitory computer-readable medium storing instructions, which when executed by a processing device, cause the processing device to perform a method comprising:
receiving, from a switch, an indication of network congestion at the switch, wherein the indication includes a contributor to the network congestion;
identifying, in response to the indication of network congestion, a host running the contributor; and
transmitting, to the host in response to the indication of network congestion, a first instruction to restrict network traffic originated by the contributor.
9. The non-transitory computer-readable medium of claim 8 , wherein the contributor is an application running on the host.
10. The non-transitory computer-readable medium of claim 8 , wherein the contributor is a data compute node running on the host.
11. The non-transitory computer-readable medium of claim 8 , the method further comprising:
receiving, from the switch, an indication of network congestion relief; and
transmitting, to the host in response to the indication of network congestion relief, a second instruction to remove the restriction on network traffic originated by the contributor.
12. The non-transitory computer-readable medium of claim 8 , wherein the restriction is a reduction in input/output (I/O) queue depth.
13. The non-transitory computer-readable medium of claim 8 , wherein the contributor to the network congestion is a largest contributor to the network congestion.
14. The non-transitory computer-readable medium of claim 8 , wherein the contributor to the network congestion is determined to contribute above a threshold amount of traffic to the network congestion.
15. A management server comprising:
a processing device; and
a memory coupled to the processing device, the memory storing instructions which, when executed by the processing device, cause the management server to:
receive, from a switch, an indication of network congestion at the switch, wherein the indication includes a contributor to the network congestion;
identify, in response to the indication of network congestion, a host running the contributor; and
transmit, to the host in response to the indication of network congestion, a first instruction to restrict network traffic originated by the contributor.
16. The management server of claim 15 , wherein the contributor is an application running on the host.
17. The management server of claim 15 , wherein the contributor is a data compute node running on the host.
18. The management server of claim 15 , wherein execution of the instructions further cause the management server to:
receive, from the switch, an indication of network congestion relief; and
transmit, to the host in response to the indication of network congestion relief, a second instruction to remove the restriction on network traffic originated by the contributor.
19. The management server of claim 15 , wherein the restriction is a reduction in input/output (I/O) queue depth.
20. The management server of claim 15 , wherein the contributor to the network congestion is a largest contributor to the network congestion or the contributor is determined to contribute above a threshold amount of traffic to the network congestion.