Intel Corporation

Robert J Woodruff

William R Magro

Jianxin Xiong

Mark Sean Hefty

Arlin R Davis

Sayantan Sur

Jerrie L Coffman

Shiow-wen Wendy Cheng

Todd M Rimmer

Frank L Berry

David M Lee

Rajesh R Shah

Roy Larsen

Sean Hefty

Shiow-Wen Cheng

William R Margo

In an embodiment, at least one interface mechanism may be provided. The mechanism may permit, at least in part, at least one process allocate, at least in part, and/or configure, at least in part, at least one network-associated object. Such allocation and/or configuration, at least in part, may be in accordance with at least one parameter set that may correspond, at least in part, to at least one query issued by the at least one process via the mechanism. Many modifications are possible without departing from this embodiment.

Discovery, configuration, and control of fabric interfaces

In an embodiment, a method is provided. In an embodiment, the method provides determining that a message has been placed in a send buffer; and transferring the message to an application on a second virtual machine by bypassing use of an operating system to process the message by directly placing the message in an application memory space from which the application can retrieve the message.

RDMA (remote direct memory access) data transfer in a virtual environment

At least one mechanism to permit, at least in part, allocation and/or configuration, at least in part, of at least one network-associated object

In an embodiment, a method is provided. In an embodiment, the method includes implementing a virtual remote direct memory access (RDMA) component in a virtualization layer on the computer system, the virtual RDMA component having an interface implementing RDMA semantics. An RDMA send request from a send queue associated with a first application running on the first VM is read using the virtual RDMA component, the RDMA send request referencing a send buffer in an application memory space for the first application. The virtual RDMA component then copies or transfers a message in the send buffer to a receive buffer in the second VM. A host computing system configured to implement the method and instructions configured to be executed on a host computing system is also provided.

Methods for performing power management of InfiniBand (IB) switches and apparatus and software configured to implement the methods. Power management datagrams (MADs) are used to inform IB switches that host servers connected to the IB switch's ports are to transition to a reduced-power or offline state or have returned to a normal operating state. A subnet management agent (SMA) on the IB switch receives the power MADs from the host servers and tracks each server's operating state. In response to power down MADs, the SMA coordinates power reduction of the switch's ports and other switch circuitry. For switches with multi-port IB interfaces, a multi-port interface is caused to enter a reduced-power state when all of its ports are connected to host servers that are idle or offline. Additionally, when all of a switch's ports are connected to idle or offline servers the SMA may put the switch's core switch logic into a reduced-power state. Power MADs are also used to inform upstream IB switches when a switch is to transition to a reduced power state or has returned to a normal operating state.

Techniques for selectively reducing power levels of ports and core switch logic in infiniband switches

Methods, apparatus, and systems for leveraging an offload programming model for local checkpoints. Compute entities in a computing environment are implemented as one or more sources and a larger number of sinks. A job dispatcher dispatches jobs comprising executable code to the source(s), and the execution of the job code is managed by the source(s). Code sections in the job code designated for offload are offloaded to the sinks by creating offload context information. In conjunction with each offload, an offload object is generated and written to storage. The offloaded code sections are executed by the sinks, which return result data to the source, e.g., via a direct write to a memory buffer specified in the offload context information. The health of the sinks is monitored to detect failures, and upon a failure the source retrieves the offload object corresponding to the code section offloaded to the failed sink, regenerates the offload context information for the code section and sends this to another sink for execution.

Leverage offload programming model for local checkpoints

Embodiments of apparatuses and methods for remote direct memory access (RDMA) with commit acknowledgements (ACKs) are described. In embodiments, a device may include a process queue to store a remote direct memory access (RDMA) request from an initiator to update a target memory. The device may further include a control module, coupled to the process queue, to issue a commit operation to the target memory based on the RDMA request and generate an acknowledgement, ACK, to be sent to the initiator on completion of the commit operation. Other embodiments may be described and/or claimed.

Apparatus and method for RDMA with commit ACKs

Methods, apparatus, systems and articles of manufacture are disclosed to manage cache memory in multi-cache environments. A disclosed apparatus includes a remote cache manager to identify a remote cache memory communicatively connected to a bus, a delegation manager to constrain the remote cache memory to share data with a host cache memory via the bus, and a lock manager to synchronize the host cache memory and the remote cache memory with a common lock state.

Methods and apparatus to manage cache memory in multi-cache environments

Methods and apparatus to provide peer-to-peer interrupt signaling between devices coupled via one or more interconnects are described. In one embodiment, a NIC (Network Interface Card such as a Remote Direct Memory Access (RDMA) capable NIC) transfers data directly into or out of the memory of a peer device that is coupled to the NIC via one or more interconnects, bypassing a host computing/processing unit and/or main system memory. Other embodiments are also disclosed.

Peer-to-peer interrupt signaling between devices coupled via interconnects

Methods, apparatus, and software for implementing RMA application programming interfaces (APIs) over Active Message (AM). AM write and AM read requests are sent from a local node to a remote node to write data to or read data from memory on the remote node using Remote Memory Access (RMA) techniques. The AM requests are handled by corresponding AM handlers, which automatically perform operations associated with the requests. For example, for AM write requests an AM write request handler may write data contained in an AM write request to a remote address space in memory on the remote node, or generate a corresponding RMA write request that is enqueued into an RMA queue used in accordance with a tagged message scheme. Similar operations are performed by AM read requests handlers. RMA reads and writes using AM are further facilitated through use of associated read, write, and RMA progress modules.

Supporting RMA API over active message

Methods and apparatus to provide heterogeneous I/O (Input/Output) using RDMA (Remote Direct Memory Access) and/or Active Message are described. In an embodiment, information is exchanged between an embedded system and a storage device via a source device. The embedded system and the storage device exchange information over a first link and a second link instead of a third link in response to a transfer rate of the first link (coupled between the embedded system and the source device) being faster than a transfer rate of the second link (coupled between the source device and the storage device). The third link is capable to directly couple the embedded system and the storage device. Other embodiments are also disclosed and claimed.

Heterogeneous input/output (I/O) using remote direct memory access (RDMA) and active message

An embodiment includes a low-latency mechanism for performing a checkpoint on a distributed application. More specifically, an embodiment of the invention includes processing a first application on a compute node, which is included in a cluster, to produce first computed data and then storing the first computed data in volatile memory included locally in the compute node; halting the processing of the first application, based on an initiated checkpoint, and storing first state data corresponding to the halted first application in the volatile memory; storing the first state information and the first computed data in non-volatile memory included locally in the compute node; and resuming processing of the halted first application and then continuing the processing the first application to produce second computed data while simultaneously pulling the first state information and the first computed data from the non-volatile memory to an input/output (IO) node.

Low latency cluster computing

Methods, apparatus, and software for optimizing network data flows within constrained systems. The methods enable data to be transferred between PCIe cards in multi-socket server platforms, each platform including a local socket having an InfiniBand (IB) HCA and a remote socket. Data to be transmitted outbound from a platform is transferred from a PCIe card to the platform's IB HCA via a proxied datapath. Data received at a platform may employ a direct PCIe peer-to-peer (P2P) transfer if the destined PCIe card is installed in the local socket or via a proxied datapath if the destined PCIe card is installed in a remote socket. Outbound transfers from a PCIe card in a local socket to the platform's IB HCA may selectively be transferred using an either a proxied data path for larger data transfers or a direct P2P datapath for smaller data transfers. The software is configured to support each of local-local, remote-local, local-remote, and remote-remote data transfers in a manner that is transparent to the software applications generating and receiving the data.

Growing community of inventors

Banks, OR, United States of America

Robert J Woodruff