International Business Machines Corporation

Globalfoundries Inc.

Maurice Isrel

Jerry Wayne Stevens

Jeffrey Douglas Haggar

Michael G Fitzpatrick

Bruce H Ratcliff

Donald W Schmidt

Hugh Edward Hockett

Constantinos Kassimis

Edward Glen Britton

Barton Clark Vashaw

Todd Elsworth Valler

Edward Zebrowski, Jr

John R Flanagan

Stephen R Valley

Lap Thiet Huynh

Benjamin P Segal

Arthur James Stagg

Mooheng Zee

Balachandar Rajaraman

Michael Jon Fox

Andrew Henry Richter

Patrick Gerald Brown

Thomas Donald Moore

Kimi A Cousins

Michael W Stayton

Daniel P Glaser

Aleksandra Brand

A host information handling system (IHS) provides virtualization of host channel adapters (HCAs). A hypervisor partitions a system memory of the host IHS into multiple logical partitions (LPARs). A particular LPAR includes a single instance of an operating system. The single instance of the operating system includes a common layer that provides virtualization of physical HCAs and sharing of the physical HCAs by multiple virtual HCAs.

Operating system virtualization for host channel adapters

A computer implemented method for receiving data to a targeted logical partition. A computer locates buffer element in reliance on a connection status bit array. The computer copies control information to the targeted logical partition's local storage. The computer updates a targeted logical partition's local producer cursor based on the control information. The computer copies data to an application receive buffer. The computer determines that an application completes a receive operation. Responsive to a determination that the application completed the receive operation, the computer a targeted logical partition's local consumer cursor to match the targeted logical partition's producer cursor.

Shared physical memory protocol

A computer implemented program product and data processing system for receiving data to a targeted logical partition. A computer locates buffer element in reliance on a connection status bit array. The computer copies control information to the targeted logical partition's local storage. The computer updates a targeted logical partition's local producer cursor based on the control information. The computer copies data to an application receive buffer. The computer determines that an application completes a receive operation. Responsive to a determination that the application completed the receive operation, the computer a targeted logical partition's local consumer cursor to match the targeted logical partition's producer cursor.

A current value of a changing operating-system-level condition can be determined, wherein the operating-system level condition is a condition of a host related to an operating system resource utilization or an operating system experienced latency. The operating-system level condition can be a condition of a layer of the OSI model above the Data Link Layer. An inbound blocking factor algorithm can execute that uses the determined current value as a variable. A value for an inbound blocking factor can be generated as a result of executing the inbound blocking factor algorithm. The generated value can be utilized as the inbound blocking factor to determine a manner in which incoming messages are to be conveyed from a network adaptor, through an adaptor interface boundary, to an operating system of the host.

Dynamically calculating an inbound blocking factor using operating-system-level conditions

A method, system and computer program product for managing at least two physical links. A TCP/IP stack configures a first interface with attributes. The TCP/IP stack activates the first interface. The TCP/IP stack detects a channel path identifier (CHPID). The TCP/IP stack creates a second interface. The TCP/IP stack associates the second interface with the first interface. The first interface and the second interface are logically linked to exchange traffic corresponding to a single internet protocol address. The TCP/IP stack configures the second interface with at least one attribute. The TCP/IP stack activates the second interface. The two physical links may rely on two different physical link types.

Transparent heterogenous link pairing

A method and system are disclosed for network adaptor optimization and interrupt reduction. The method may also build an outbound buffer list based on outgoing data and add the outgoing data to an outbound buffer queue. Furthermore, the method may set a buffer state from an empty state to a primed state to indicate that the outgoing data is prepared for transmitting and signal a network adaptor with a notification signal.

Network adaptor optimization and interrupt reduction

A system and computer program product for managing at least two physical links. A TCP/IP stack configures a first interface with attributes. The TCP/IP stack activates the first interface. The TCP/IP stack detects a channel path identifier (CHPID). The TCP/IP stack creates a second interface. The TCP/IP stack associates the second interface with the first interface. The first interface and the second interface are logically linked to exchange traffic corresponding to a single internet protocol address. The TCP/IP stack configures the second interface with at least one attribute. The TCP/IP stack activates the second interface. The two physical links may rely on two different physical link types.

A method and system are disclosed for network adaptor optimization and interrupt reduction. The method may generate an immediate I/O interrupt notification to a host device driver in response to receiving data from a network. The method may also update an inbound buffer list based on the received data in the inbound buffer queue. Furthermore, the method may set the buffer state from an empty state to a primed state to indicate that the received data is available for processing. The method may also build an outbound buffer list based on outgoing data and add the outgoing data to an outbound buffer queue. Furthermore, the method may set a buffer state from an empty state to a primed state to indicate that the outgoing data is prepared for transmitting and signal a network adaptor with a notification signal.

A method, apparatus, and computer implemented instructions for managing traffic in a network data processing system. Traffic for a plurality of network paths is monitored. In response a packet for a particular network path within the plurality of network paths causing traffic for the particular network path to exceed a level of traffic allowed, an amount of available bandwidth is reduced based on a fair share for the particular network path.

Method and apparatus for managing aggregate bandwidth at a server

The present invention provides a system and method for controlling tracing functions on a remote device. The system includes means for receiving a diagnostic signal from a host device on the remote device and means for associating the diagnostic signal with a tracing state of the remote device. In addition, the system includes means for performing a tracing function indicated by the diagnostic signal. The present invention can also be viewed as a method for controlling tracing functions on a remote device. The method operates by receiving a diagnostic signal from a host device on the remote device and associating the diagnostic signal with a tracing state of the remote device. Then, the method performs a tracing function indicated by the diagnostic signal.

Apparatus and method to integrate hardware adapter tracing with a host OS tracing through signaling

The present invention provides a system and method for controlling diagnostic functions on a remote device. The system includes a host, and a network adapter in communication with the host. The system performs a method comprising receiving a diagnostic signal from the host device on the remote device, associating the diagnostic signal with a diagnostic state of the remote device, and performing a diagnostic function indicated by the diagnostic signal.

Apparatus and method to integrate hardware adapter diagnostics with a host OS diagnostics through signaling

The invention ensures that a single and consistent reply is made to ARP request in a system of connected IP networks. In each host, the first adapter A to become active is designated as being in an arbitrary first physical network (PNET ). For each subsequent adapter B to become active on a host, the host sends a packet over one of the adapters (A) of each network already known to the host with a hop count of one. If the packet is received over adapter B, it is known that adapters A and B are in the same physical network. If this occurs, adapter B is marked as being in the same network PNET as adapter A. If the packet is not received over adapter B then B is in a different physical network as A. In this event, adapter B is marked as being in a new network PNET . When adapter A becomes inactive, if there are other active adapters in the physical network to which A belongs, then one of the remaining adapters B in that physical network is designated to have the responsibility for replying to ARP requests for A. A gratuitous ARP advertisement is transmitted into the network mapping IP-A with MAC-B. When adapter A again becomes active, adapter A will re-assume ownership of ARP replies for IP address A.

Method and apparatus for generating replies to address resolution protocol requests

The invention ensures that a single and consistent reply is made to an ARP request in a system of connected IP networks. When an adapter becomes active, the relative network on which it resides is determined by transmitting control packets over it and all other adapters known to the host and observing if and where responses are returned to the adapters. One adapter on a network is designated as active. If the same network contains other adapters, they are marked as standby adapters for the purpose of responding to ARP messages. Special processing is provided for offload adapters that perform there own ARP processing.

Growing community of inventors

Raleigh, NC, United States of America

Maurice Isrel