Method to prevent I/O starvation of slower hosts in a multi-host environment

摘要

A method, system and computer-usable medium are disclosed for managing storage control unit input/output (I/0) operations in a multi-host environment. A storage control unit sends a busy status message when it has exhausted its resources to process new I/O requests. Once the utilization of the storage control unit's resources has subsided to a predetermined upper utilization value, it sends an availability status message to a first predetermined number of host systems that are owed an availability status message. After the first predetermined number of availability status messages are sent, the storage control unit then waits for a predetermined period of time before sending a second predetermined number of availability status messages.

主权项

1. A computer-implemented method for managing storage control unit input/output (I/O) operations in a multi-host environment, comprising:
generating a set of busy status messages in response to receiving a corresponding set of I/O requests generated by an associated set of hosts, the set of busy status messages generated by a storage control unit when it has insufficient resources to process the set of I/O requests, each of the set of busy status messages comprising a CUBusy message, the CUBusy message being defined by a Fibre Connection protocol; generating a set of availability status messages when the utilization of the storage control unit's resources subsides to a first utilization value, the first utilization value representing a percentage of resources of the storage control unit, the set of availability status messages corresponding to the set of busy status messages, each of the set of availability status messages comprising a CUEnd message, the CUEnd message being defined by the Fibre Connection protocol; providing a first subset of the set of availability status messages to a first subset of the set of hosts; and providing a second subset of the set of availability status messages to a second subset of the set of hosts, the provision of the second subset of the set of availability status messages delayed until a predetermined time interval has elapsed.