Abstract: A method and apparatus are provided for implementing system to system communication in a switchless non-InfiniBand (IB) compliant environment. IB architected multicast facilities are used to communicate between HCAs in a loop or string topology. Multiple HCAs in the network subscribe to a predetermined multicast address. Multicast messages sent by one HCA destined to the pre-determined multicast address are received by other HCAs in the network. Intermediate TCA hardware, per IB architected multicast support, forward the multicast messages on via hardware facilities, which do not require invocation of software facilities thereby providing performance efficiencies. The messages flow until picked up by an HCA on the network. Architected higher level IB connections, such as IB supported Reliable Connections (RCs) are established using the multicast message flow, eliminating the need for an IB Subnet Manager (SM).

Abstract: An apparatus and method allows processing sequenced records across multiple network connections. A “logical connection” is defined to include one or more network connections. Each message is assigned a sequence number that allows the messages to be ordered on the other end according to sequence number, regardless of which network connection in the logical connection is used to transfer the message. By defining messages, sequencing those messages, and transferring the messages over multiple network connections, the throughput and performance of networked computer systems are substantially increased.

Abstract: In a clustered computer system with multiple power domains, a bus number manager within each power domain manages multiple nodes independently of other power domains. A node within a specified power domain includes a non-volatile memory that includes bus numbering information for its own buses as well as bus numbering information for two of its logically-interconnected neighbors. This creates a distributed database of the interconnection topology for each power domain. Because a node contains bus numbering information about its logical neighbor node(s), the bus numbers for the buses in the nodes are made persistent across numerous different system reconfigurations. The clustered computer system also includes a bus number manager that reads the non-volatile memories in the nodes during initial program load (i.e.

Abstract: A method and apparatus are provided for implementing system to system communication in a switchless non-InfiniBand (IB) compliant environment. IB architected multicast facilities are used to communicate between HCAs in a loop or string topology. Multiple HCAs in the network subscribe to a predetermined multicast address. Multicast messages sent by one HCA destined to the pre-determined multicast address are received by other HCAs in the network. Intermediate TCA hardware, per IB architected multicast support, forward the multicast messages on via hardware facilities, which do not require invocation of software facilities thereby providing performance efficiencies. The messages flow until picked up by an HCA on the network. Architected higher level IB connections, such as IB supported Reliable Connections (RCs) are established using the multicast message flow, eliminating the need for an IB Subnet Manager (SM).

Abstract: Provided are a method, system, and program for configuring shared devices over a fabric. A module in a first processing complex configures a first part of a fabric enabling communication with a set of devices accessible through the first part of the fabric. The module detects a located device accessible through a second part of the fabric, wherein a second processing complex is designated to configure the second part of the fabric and the located device. The module determines whether the second processing complex is available in response to detecting the uninitialized device. The module passes to a device driver in the first processing complex an uninitialized property for the located device. The device driver requests the module to configure the second part of the fabric to enable access to the located device over the second part of the fabric in response to determining that the located device has the uninitialized property.

Abstract: In a clustered computer system with multiple power domains, a bus number manager within each power domain manages multiple nodes independently of other power domains. A node within a specified power domain includes a non-volatile memory that includes bus numbering information for its own buses as well as bus numbering information for two of its logically-interconnected neighbors. This creates a distributed database of the interconnection topology for each power domain. Because a node contains bus numbering information about its logical neighbor node(s), the bus numbers for the buses in the nodes are made persistent across numerous different system reconfigurations. The clustered computer system also includes a bus number manager that reads the non-volatile memories in the nodes during initial program load (i.e.

Abstract: An apparatus, program product and method support the dynamic modification of cluster communication parameters such as a fragmentation size parameter through controllably deferring the processing of a requested fragmentation size change in a source node until after receipt an acknowledgment message for at least one unacknowledged message sent by the source node to a plurality of target nodes. By controllably deferring such processing until it is confirmed that any such previously-unacknowledged messages sent by a source node have been received by any target nodes, synchronization between the source node and the target nodes may be obtained, and a fragmentation size change may occur in a coordinated fashion such that future messages from the source node to the target node will be processed by both the source and the target nodes using the modified fragmentation size parameter.

Abstract: An apparatus, program product and method utilize cluster data port services within a cluster infrastructure to provide reliable and efficient communications between nodes in a clustered computer system. The cluster data port services present an abstracted transport service that encapsulates and manages the establishment of multiple connection paths between a source node, a target node and one or more backup nodes in such a manner that a cluster data port is effectively utilized as single data port from the perspective of a user program.

Abstract: An apparatus and method allows processing sequenced records across multiple network connections. A “logical connection” is defined to include one or more network connections. Each message is assigned a sequence number that allows the messages to be ordered on the other end according to sequence number, regardless of which network connection in the logical connection is used to transfer the message. By defining messages, sequencing those messages, and transferring the messages over multiple network connections, the throughput and performance of networked computer systems are substantially increased.

Abstract: A method, apparatus and computer program product implement InfiniBand (IB) network topology simplification. A Subnet Manager (SM) of an IB subnet sends a subnet discovery request to each switch requesting the number of ports that are attached to the switch. Each of the switches and target channel adapters (TCAs) within the IB subnet includes a Subnet Management Agent (SMA). The Subnet Management Agent (SMA) of the receiving switch responds to the SM indicating a sufficient number of ports on the switch to support at least one port for each TCA. Each TCA supports at least two local IDs (LIDs).

Abstract: An apparatus and method allows processing sequenced records across multiple network connections. A “logical connection” is defined to include one or more network connections. Each message is assigned a sequence number that allows the messages to be ordered on the other end according to sequence number, regardless of which network connection in the logical connection is used to transfer the message. By defining messages, sequencing those messages, and transferring the messages over multiple network connections, the throughput and performance of networked computer systems are substantially increased.

Abstract: A method for assigning an identifier to data processed through protocol layers in one or more computers over a network. A space for the identifier is reserved in the header of each protocol layer. The identifier is then generated at one of the protocol layers. In an embodiment, the identifier is generated at the lowest protocol layer of a computer that sends the data, i.e., the sending computer. Once the identifier is generated, it is then stored in the reserved space in the header.

Abstract: Methods, systems and articles of manufacture for automatically starting a node in a clustered computer system. A starting state value may be assigned to the node and a discovery process initiated to find a sponsor node. If a sponsor node is found, the node is joined with the sponsor node in the clustered computer system. If a sponsor node is not found, the node is started as a one-node cluster in the clustered computer system. An active state value is assigned to the node upon inclusion into the clustered computer system.

Abstract: A clustered computer system includes multiple computer systems (or nodes) coupled together via one or more networks that can become members of a group to work on a particular task. Each node includes a cluster engine, a cluster communication mechanism that includes a sliding send window, and one or more service tasks that process messages. The sliding send window allows a node to send out multiple messages without waiting for an individual acknowledgment to each message. The sliding send window also allows a node that received the multiple messages to send a single acknowledge message for multiple received messages. By using a sliding send window to communicate with other computer systems in the cluster, the communication traffic in the cluster is greatly reduced, thereby enhancing the overall performance of the cluster. In addition, the latency between multiple messages sent concurrently is dramatically reduced.

Abstract: A method and system for communication in a system area network (SAN) data processing system are described. The SAN includes a plurality of interconnected nodes that each have at least one port for communication. To avoid communication-induced errors that may arise, for example, if multiple nodes share the same node ID, the port of a node in the SAN is marked as “fenced” to prevent transmission of packets of a first traffic type while permitting transmission of packets of a second traffic type. The marking of the port may be recorded, for example, in a configuration register of the port. While the port is fenced, only packets of other than the first traffic type are routed via the port. In one preferred embodiment, the second traffic type represents SAN configuration traffic, and the first traffic type represents non-configuration traffic.

Abstract: A method and system for communication in a system area network (SAN) data processing system are described. The SAN includes a plurality of interconnected nodes that each have at least one port for communication. To avoid communication-induced errors that may arise, for example, if multiple nodes share the same node ID, the port of a node in the SAN is marked as “fenced” to prevent transmission of packets of a first traffic type while permitting transmission of packets of a second traffic type. The marking of the port may be recorded, for example, in a configuration register of the port. While the port is fenced, only packets of other than the first traffic type are routed via the port. In one preferred embodiment, the second traffic type represents SAN configuration traffic, and the first traffic type represents non-configuration traffic.

Abstract: A method, apparatus, system, and signal-bearing medium that in an embodiment dynamically allocate and/or deallocate data structures on demand to respective partitions in a logically-partitioned electronic device. The data structures are associated with an adapter, and the partitions may use the data structures to access the adapter and, in an embodiment, to send and/or receive messages across a network.

Abstract: According to the present invention, a communications protocol supporting cluster configurations more complex than a single LAN is disclosed. A cluster destination address table (CDAT) is used in conjunction with a network message servicer to communicate between computer systems in a cluster. Each computer system preferably contains a cluster servicer, a CDAT, and a network message servicer. The CDAT contains network addresses, status and adapter information for each computer system in a cluster. Although computer systems may have alternate network addresses when they have multiple adapters, the CDAT indexes primary and alternate address information under a single named system. Thus, redundant connections amongst computer systems are identified, while still using the numeric addresses upon which the network message servicer is based. To send a message using the methods of the present invention, the cluster servicer retrieves a network address for a computer system from a CDAT.

Abstract: An apparatus, program product and method support the dynamic modification of cluster communication parameters through a distributed protocol whereby individual nodes locally confirm initiation and status information for every node participating in a parameter modification operation. By doing so, individual nodes are also able to locally determine the need to undo locally-performed parameter modifications should any other node be incapable of performing a parameter modification. Moreover, specifically with respect to cluster communication parameters such as heartbeat parameters, such parameters may be dynamically modified by configuring a sending node to send a heartbeat message to a receiving node, with the heartbeat message indicating that a heartbeat parameter is to be modified. In response to the heartbeat message, the receiving node may then send an acknowledgment message to the sending node that indicates whether the heartbeat parameter has been modified in the receiving node.

Abstract: A method and system for communication in a system area network (SAN) data processing system are described. The SAN includes a plurality of interconnected nodes that each have at least one port for communication. To avoid communication-induced errors that may arise, for example, if multiple nodes share the same node ID, the port of a node in the SAN is marked as “fenced” to prevent transmission of packets of a first traffic type while permitting transmission of packets of a second traffic type. The marking of the port may be recorded, for example, in a configuration register of the port. While the port is fenced, only packets of other than the first traffic type are routed via the port. In one preferred embodiment, the second traffic type represents SAN configuration traffic, and the first traffic type represents non-configuration traffic.