Manageability tools for lossless networks

Abstract

Manageability tools are provided for allowing an administrator to have better control over switches in a lossless network of switches. These tools provide the ability to detect slow drain and congestion bottlenecks, detect stuck virtual channels and loss of credits, while hold times on edge ASICs to be different from hold times encore ASICs, and mitigate severe latency bottlenecks.

What is claimed is: 
     
         1 . A network switch, comprising:
 a first port, configured for transmitting frames of data; and   a maximum time value that a frame may wait on the network switch before being dropped, having a default value;   logic to mitigate a slow-drain bottleneck on the first port by timing out frames waiting for the first port;   logic to adjust the maximum time value to a first value if the network switch is an edge switch; and   logic to adjust the maximum time value to a second value if the network switch is not an edge switch.   
     
     
         2 . The network switch of  claim 1 , wherein the first value is less than default value. 
     
     
         3 . The network switch of  claim 1 , wherein the second value is greater than the default value. 
     
     
         4 . The network switch of  claim 1 , wherein the logic to mitigate a slow-drain bottleneck on the first port is implemented in software. 
     
     
         5 . The network switch of  claim 1 , wherein the logic to mitigate a slow-drain bottleneck on the first port is implemented in hardware. 
     
     
         6 . The network switch of  claim 1 , further comprising:
 a first maximum time value that a frame may wait for transmission on the first port, wherein frames that wait for transmission by the first port longer than the first maximum time value are dropped;   a second port, configured for transmitting frames of data; and   a second maximum time value that a frame may wait for transmission on the second port, wherein frames that wait for transmission by the second port longer than the second maximum time value are dropped, and   wherein the logic to mitigate a slow-drain bottleneck on the first port comprises:   logic to adjust the first maximum time value independently of the second maximum time value.   
     
     
         7 . The network switch of  claim 1 , further comprising:
 a first maximum time value that a frame may wait for transmission on a first virtual channel of the first port, wherein frames that wait for transmission on the first virtual channel longer than the first maximum time value are dropped; and   a second maximum time value that a frame may wait for transmission on a second virtual channel of the first port, wherein frames that wait for transmission on the second virtual channel longer than the second maximum time value are dropped, and wherein the logic to mitigate a slow-drain bottleneck on the first port comprises:   logic to adjust the first maximum time value without changing the second maximum time value.   
     
     
         8 . A method comprising:
 mitigating a slow-drain bottleneck in a network switch by timing out frames waiting on the network switch, comprising:   configuring a timeout value for the network switch to a default timeout value, wherein the network switch drops frames that wait for transmission on the network switch longer than the timeout value; and   adjusting the timeout value based on whether the network switch is an edge switch.   
     
     
         9 . The method of  claim 8 , wherein adjusting the timeout value based on whether the network switch is an edge switch comprises:
 reducing the timeout value if the network switch is an edge switch.   
     
     
         10 . The method of  claim 8 , wherein adjusting the timeout value based on whether the network switch is an edge switch comprises:
 increasing the timeout value if the network switch is not an edge switch.   
     
     
         11 . The method of  claim 8 , wherein mitigating a slow drain bottleneck comprises:
 defining a first timeout value that a frame may wait for transmission on a first port of the network switch, wherein the network switch drops frames that wait for transmission by the first port longer than the first timeout value;   defining a second timeout value that a frame may wait for transmission on a second port of the network switch, wherein the network switch drops frames that wait for transmission by the second port longer than the second timeout value, wherein the second timeout value is independent of the first timeout value; and   adjusting the first timeout value.   
     
     
         12 . The method of  claim 8 , wherein mitigating a slow drain bottleneck comprises:
 defining a first timeout value that a frame may wait for transmission on a first virtual channel of a first port of the network switch, wherein the network switch drops frames that wait for transmission on the first virtual channel longer than the first timeout value;   defining a second timeout value that a frame may wait for transmission on a second virtual channel of the first port, wherein the network switch drops frames that wait for transmission on the second virtual channel of the first port longer than the second timeout value, wherein the second timeout value is independent of the first timeout value; and   adjusting the first timeout value.   
     
     
         13 . A non-transitory computer readable medium, on which instructions for execution by network switch are stored, the instructions comprising instructions that when executed cause the network switch to:
 mitigate a slow drain bottleneck by timing out frames waiting on the network switch;   configure a timeout value for the network switch to a default timeout value, wherein the network switch drops frames that wait for transmission on the network switch longer than the timeout value; and   adjust the timeout value based on whether the network switch is an edge switch.   
     
     
         14 . The computer readable medium of  claim 13 , wherein the instructions that when executed cause the network switch to adjust the timeout value comprise instructions that when executed cause the network switch to
 decrease the timeout value if the network switch is an edge switch.   
     
     
         15 . The computer readable medium of  claim 13 , wherein the instructions that when executed cause the network switch to adjust the timeout value comprise instructions that when executed cause the network switch to
 increase the timeout value if the network switch is not an edge switch.   
     
     
         16 . The computer readable medium of  claim 13 , wherein the instructions further comprise instructions that when executed cause the network switch to:
 adjust a maximum time that a frame may wait for transmission on a virtual channel of a port of the network switch.   
     
     
         17 . The computer readable medium of  claim 13 , wherein the instructions further comprise instructions that when executed cause the network switch to:
 adjust a maximum time that a frame may wait for transmission on a port of the network switch.   
     
     
         18 . The computer readable medium of  claim 13 , wherein the instructions further comprise instructions that when executed cause the network switch to:
 report adjustment of the timeout value.   
     
     
         19 . The computer readable medium of  claim 13 , wherein the instructions that when executed cause the network switch to adjust the timeout value comprise instructions that when executed cause the network switch to:
 adjust a first timeout value for a first network switch element; and   adjust a second timeout value for a second network switch element, independently of the first timeout value.