Using machine learning to detect slow drain conditions in a storage area network
In one embodiment, a device obtains one or more fabric port (F-port) counters and one or more extender port (E-port) counters in a storage area network (SAN). The device inputs the obtained F-port and E-port counters to a machine learning-based prediction model. The device uses the prediction model to predict a slow drain condition in the SAN, based on the counters input to the model. The device initiates a corrective measure in the SAN, based on the predicted slow drain condition.
1. A method comprising:
obtaining, by a device, one or more fabric port (F-port) counters and one or more extender port (E-port) counters in a storage area network (SAN);
inputting, by the device, the obtained F-port and E-port counters to a machine learning-based prediction model;
using, by the device, the prediction model to predict a slow drain condition in the SAN, based on the counters input to the model; and
initiating, by the device, a corrective measure in the SAN, based on the predicted slow drain condition.
2. The method as in claim 1 , wherein initiating the corrective measure in the SAN comprises at least one of: initiating a port quarantine, a port reset, or a port ingress rate limiter in the SAN.
3. The method as in claim 1 , further comprising:
obtaining, by the device, flow level metrics for an F-port in the SAN; and
using, by the device, the flow level metrics as input to a classifier, to predict the slow drain condition.
4. The method as in claim 3 , wherein the classifier comprises a multiple logistic regression model.
5. The method as in claim 1 , further comprising:
training the prediction model in part by selecting which F-port counters are most predictive of the one or more E-port counters.
6. The method as in claim 1 , wherein the prediction model is a multiple linear regression model.
7. The method as in claim 1 , wherein the counters are indicative of at least one of: a frame drop count, a data rate, a transmit wait count, or a port credit-related counter.
8. The method as in claim 1 , wherein the slow drain condition is attributable to oversubscription.
9. An apparatus, comprising:
one or more network interfaces to communicate with a network;
a processor coupled to the network interfaces and configured to execute one or more processes; and
a memory configured to store a process executable by the processor, the process when executed configured to:
obtain one or more fabric port (F-port) counters and one or more extender port (E-port) counters in a storage area network (SAN);
input the obtained F-port and E-port counters to a machine learning-based prediction model;
use the prediction model to predict a slow drain condition in the SAN, based on the counters input to the model; and
initiate a corrective measure in the SAN, based on the predicted slow drain condition.
10. The apparatus as in claim 9 , wherein the apparatus initiates the corrective measure in the SAN by at least one of: initiating a port quarantine, a port reset, or a port ingress rate limiter in the SAN.
11. The apparatus as in claim 9 , wherein the process when executed is further configured to:
obtain flow level metrics for an F-port in the SAN; and
use the flow level metrics as input to a classifier, to predict the slow drain condition.
12. The apparatus as in claim 11 , wherein the classifier comprises a multiple logistic regression model.
13. The apparatus as in claim 9 , wherein the process when executed is further configured to:
train the prediction model in part by selecting which F-port counters are most predictive of the one or more E-port counters.
14. The apparatus as in claim 9 , wherein the prediction model is a multiple linear regression model.
15. The apparatus as in claim 9 , wherein the counters are indicative of at least one of: a frame drop count, a data rate, a transmit wait count, or a port credit-related counter.
16. The apparatus as in claim 9 , wherein the slow drain condition is attributable to oversubscription.
17. A tangible, non-transitory, computer-readable medium storing program instructions that cause a device to execute a process comprising:
obtaining, by the device, one or more fabric port (F-port) counters and one or more extender port (E-port) counters in a storage area network (SAN);
inputting, by the device, the obtained F-port and E-port counters to a machine learning-based prediction model;
using, by the device, the prediction model to predict a slow drain condition in the SAN, based on the counters input to the model; and
initiating, by the device, a corrective measure in the SAN, based on the predicted slow drain condition.
18. The computer-readable medium as in claim 17 , wherein initiating the corrective measure in the SAN comprises at least one of: initiating a port quarantine, a port reset, or a port ingress rate limiter in the SAN.
19. The computer-readable medium as in claim 17 , wherein the counters are indicative of at least one of: a frame drop count, a data rate, a transmit wait count, or a port credit-related counter.
20. The computer-readable medium as in claim 17 , wherein the prediction model is a multiple linear regression model.