International Business Machines Corporation

Alison B White

Aruna V Ramanan

Casimer M DeCusatis

Edward Joseph Seminaro

Daniel Guy Young

Robert Sidney Blackmore

Richard R Treumann

Nicholas P Rash

John Dorfner

Karen F Rash, Legal Representative

Casandra H Qiu

Embodiments relate to providing communication over cross-coupled links between independently managed compute and storage networks. An aspect includes coupling an independently managed local subsystem with an independently managed remote subsystem over cross-coupled links, whereby each subsystem includes compute entities and storage entities. Unique identifiers are assigned to all the compute entities and the storage entities in the local network and the remote network. A determination is then made as to whether each entity is in the local subsystem or the remote subsystem. Accordingly, a global broadcast tree is built to bridge the compute entities in the local subsystem to the storage entities in both the local and remote subsystem. Responsive to an error in a layer of the local subsystem external to a cross-coupled link, the cross-coupled link in the local subsystem is disabled. Accordingly, the remote subsystem may detect that the link has failed.

Enabling communication over cross-coupled links between independently managed compute and storage networks

A method of validating a configuration of a computer clusters includes transmitting a first neighbor identification to a first flexible service processor (FSP) arranged in the first computer cluster and a second neighbor identification to a second FSP arranged in the second computer cluster, connecting a first end of a cable to a first transceiver arranged in the first cluster and connecting a second end of the cable to a second transceiver arranged in the second cluster. The first neighbor identification is passed from the first transceiver to the second computer cluster and the second neighbor identification is passed from the second transceiver toward the first computer cluster. The first neighbor identification is compared with a desired first neighbor identification to establish a first comparison result, and the second neighbor identification is compared with a desired second neighbor identification to establish a second comparison result and a notice is generated.

Configuration validation system for computer clusters

A method of validating multi-cluster computer interconnects includes calculating a cable interconnect table associated with the multi-cluster computer, and distributing the cable interconnect table to a first transceiver in the first computer cluster and a second transceiver in the second computer cluster. The method also includes connecting a first end of a cable to the first transceiver and a second end of the cable to the second transceiver, transmitting a first neighbor identification from the first cluster to the second cluster, and a second neighbor identification from the second cluster to the first cluster, comparing the first neighbor identification with a desired first neighbor identification from the cable interconnect table to establish a first comparison result and the second neighbor identification with a desired second identification from the cable interconnect table to establish a second comparison result, and generating an alert based on the first and second comparison results.

Validation of computer interconnects

Endpoint identifiers are dynamically assigned to network interfaces, in response to a change in physical connection. When a physical connection associated with a network interface is changed, such as a cable coupling the network interface to an endpoint is moved from one location to another location, a new endpoint identifier is assigned to the network interface. The new endpoint identifier is based on the location of the new endpoint.

Dynamically assigning endpoint identifiers to network interfaces of communications networks

A method of topology discovery and identification of switches enables a user to determine the topology of a N-stage switch network. The method includes ascertaining an intended topology of the N-stage switch network, creating a list of switch boards present in the N-stage switch network, and determining a switch board connection pattern. The method further includes classifying each of the switch boards as an outer switch board, or an inner switch board, and creating a list for each type of switch board, classifying each of the switch boards on the OB list as stage 1, stage 2, or unknown, classifying each of the switch boards on the IB list as stage 3 to stage N, grouping the stage 2 and stage 1 switch boards into sectors, and numbering each type of switch board, thereby obtaining a determined topology, and validating the determined topology by comparing it to the intended topology.

Topology discovery and identification of switches in an N-stage interconnection network

Topology discovery and identification of switches enables a user to determine the topology of a three-stage switch network. The method includes ascertaining an intended topology of the three-stage switch network, creating a list of switch boards that are present in the three-stage switch network, and determining a switch board connection pattern by obtaining information indicating how each switch board is connected to neighboring switch boards. The method further includes classifying each of the switch boards on the list of switch boards as a node switch board, an intermediate switch board, or a jump switch board, and creating a list for each type of switch board. The method further includes grouping the intermediate and node switch boards into sectors, and numbering each type of switch board, thereby obtaining a determined topology, and validating the determined topology by comparing it to the intended topology.

Topology discovery and identification of switches in a three-stage interconnection network

A deterministic, non-deadlocking technique to achieving distributed consensus in a multithreaded multiprocessing computing environment is provided. A communicator is established across multiple processes in the multithreaded computer environment notwithstanding that multiple groups of threads may be simultaneously trying to establish communicators. The technique includes communicating across the multiple processes to establish a candidate identifier for the communicator for a group of participating threads of the multiple processes; and communicating across the multiple processes to check at each participating thread of the multiple processes whether the candidate identifier can be claimed at its process, and if so, claiming the candidate identifier as the new identifier thereby establishing the communicator. As one example, the technique can be implemented via a subroutine call within a message passing interface (MPI) library.

Growing community of inventors

Kingston, NY, United States of America

Alison B White