Abstract: A technique is provided for assigning an I/O controller to a host in a cluster. The cluster includes one or more hosts and one or more I/O controllers connected by a cluster interconnection fabric. In an example embodiment, an I/O controller is connected to the cluster interconnection fabric. The I/O controller connected to the fabric is detected and a network address is assigned to the I/O controller. An administrative agent is used to assign the I/O controller to a host that is connected to the cluster interconnection fabric. A message is sent to the host informing the host that the I/O controller is assigned to the host and providing the network address of the I/O controller.

Abstract: A method for causing a host channel adaptor which is involved with a clustered arrangement to hibernate. Before the HCA can hibernate, it is necessary for its clients to hibernate first. Once this is accomplished, all data is stored in memory and the HCA goes into hibernation. It resumes operation when a request is received. The HCA is checked to see if it has been changed and various parameters are examined to determine if an error has occurred which is unrecoverable. If not, the operation of the device is resumed.

Abstract: A cluster includes a fabric, a host coupled to the fabric, an I/O enclosure coupled to the fabric, and a subnet manager. The subnet manager performs a topology discovery of the cluster and detects ports associated with the fabric. The subnet manager computes a minimal spanning tree for the cluster, which connects every port to every other port through a single path. The subnet manager assigns a single base LID to each port and programs the assigned LIDs into forwarding tables in the fabric. Additional LIDs are reserved for each port while the single base LID is assigned to each port such that the fabric is functional and connected when the subnet manager is performing a path analysis of the cluster.

Abstract: A topology change notification mechanism is provided to notify topology changes in a subnet of a switched fabric including at least a host system, a target system and switches interconnected via links. Such a mechanism may be installed in a host system to allow a client at one of the host system and the target system to create and communicate a list of topology changes that are interesting to the client for topology change notifications; determining if a topology change occurred in the switched fabric is in the list of topology changes created by the interested client; and reporting a topology change event to the interested client if the topology change is in the list of topology changes created by the interested client.

Abstract: A host is coupled to a cluster interconnection fabric which includes a fabric-attached I/O controller. The host includes a processor, a memory coupled to the processor and an operating system. The operating system includes a kernel and a fabric bus driver to provide an I/O bus abstraction to the kernel for the cluster interconnection fabric. The fabric bus driver presents the cluster interconnection fabric to the kernel as a local I/O bus, and presents the fabric-attached I/O controller to the kernel as a local I/O controller attached to a local I/O bus.

Abstract: A host is coupled to a cluster interconnection fabric which includes a fabric-attached I/O controller. The host includes a processor, a memory coupled to the processor and an operating system. The operating system includes a kernel and a fabric bus driver to provide an I/O bus abstraction to the kernel for the cluster interconnection fabric. The fabric bus driver presents the cluster interconnection fabric to the kernel as a local I/O bus, and presents the fabric-attached I/O controller to the kernel as a local I/O controller attached to a local I/O bus.

Abstract: A cluster operating in accordance with an integrating operating system independent power management with operating system directed power management includes a group of hosts connected together by a cluster interconnection fabric. A cluster administrator is connected to the group of hosts via the fabric and the cluster administrator includes a cluster power manager. A group of input/output units are connected to the group of hosts and the cluster interconnection fabric. Each of the hosts includes a controller element and an operating system power manager and input/output controller device driver stack. The cluster administrator transmits a request to the controller element of one of the hosts via the fabric and receives a reply therefrom and transmits a command. The controller element transmits the command to the operating system power manager and the input/output controller device driver stack of its host and transmits a command completion acknowledgment to the cluster power manager.

Abstract: A mechanism is provided at a host node to determine the optimal number of receive buffers to post for entities (clients) in order to receive incoming data messages in a switched fabric for scalable solutions.

Abstract: A mechanism is provided at a host system to allow multiple entities (clients) to send and receive messages of a particular class of management services in a switched fabric for scalable solutions.

Abstract: A dynamic workload feedback mechanism is provided at a service provider to notify a client a current workload of the service provider in a switched fabric in order to prevent unnecessary timeouts and retries of duplicate service requests.