International Business Machines Corporation

Kavitha Vittal Murthy Baratakke

Venkat Venkatsubra

Vinit Jain

Nikhil Hegde

Vasu Vallabhaneni

Dwip N Banerjee

Jos Manuel Accapadi

Jeffrey Paul Messing

Rashmi Narasimhan

Andrew Dunshea

Shaival Jagdishbhai Chokshi

Lilian Sylvia Fernandes

Vishal Ramachandra Mansur

Ketan Priyakant Pancholi

Adekunle Bello

Aruna Yedavilli

Deanna Lynn Quigg Brown

Sivakumar Krishnasamy

Xiaohan Qin

Bernard A King-Smith

Pramod Achyut Bhandiwad

Nikhil Hedge

G Shantala

Varun Sethi

Rashmi Narashimhan

Systems, methods, and computer program products to perform an operation comprising providing a plurality of assist threads configured to process data units received by a network adapter, wherein each of the plurality of assist threads comprises a queue configured to store data units allocated to the respective assist thread, allocating data units received by the network adapter to assist threads in a subset of the plurality of assist threads according to a first function, wherein the subset includes a first assist thread and does not include a second assist thread, of the plurality of assist threads, monitoring the queues of the assist threads in the subset, and upon determining that a count of data units in the queue of the first assist thread exceeds a first threshold, modifying the first function to include the second assist thread in the subset of the plurality of assist threads.

Dynamically optimized device driver protocol assist threads

Provided are techniques for active SEA learning about a client LPAR MAC addresses via address resolution protocol (ARP) packets received on a virtual interface (of the active SEA). Any new client MAC addresses learned on the active SEA are sent to the inactive SEA via a control channel. When SEA failover happens, as the previously inactive SEA is about to become active, it will first send out RARP (reverse ARP) packets with the client MAC addresses as the source MAC addresses respectively. This effectively informs the switch connected to the previously inactive SEA that these client MAC addresses are to be routed through this switch port; the client MAC addresses saved on the switch connected to the previously active SEA are cleared as a result.

SEA failover mechanism with minimized packet losses

In a method for regulating network bandwidth in a virtualized computer environment, a computer having a hypervisor program receives a request from a first virtual client to transmit data. In response, the computer transfers the data from a memory of the first virtual client to a memory of a virtual server. The computer receives an error notification from a shared virtual network adapter of the virtual server, indicative of insufficient network bandwidth available to transmit the data. In response, the computer notifies the first virtual client that insufficient network bandwidth is available to transmit the data.

Regulating network bandwidth in a virtualized environment

In a method for managing shared Ethernet adapters in a virtualized computer environment, a computer programmed with a hypervisor program receives a first request from a virtual client to transmit data. In response to receiving the first request, the computer programmed with the hypervisor program selects a first active virtual server from a first plurality of concurrently active virtual servers of the virtualized computer environment. The computer programmed with the hypervisor program copies the data from a memory of the virtual client to a trunk adapter of a first shared Ethernet adapter of the selected first active virtual server.

Aggregating shared Ethernet adapters in a virtualized environment

A method, system, and computer program product for a self-adjusting application heartbeat management process are provided in the illustrative embodiments. A first application receives from a second application, a response message in response to a heartbeat message. Information is identified in the response message usable for computing a delay in a future response message for a future heartbeat message, such that the information communicates to the first application a system condition existing in the second data processing system where the second application is executing. A response time is adjusted using the information to form a modified response time. The future response message is received after the response time.

Self-adjusting application heartbeat management

An approach is provided to improve network efficiency. A send segment size, such as a maximum segment size (MSS), that corresponds to data segments being sent to a receiver over a computer network. A data block is identified in a send buffer and the data block includes more than one data segments. Based on the determined send segment size, all but a remaining data segment of the data segments are sent to the receiver. The sent data segments are each the determined send segment size and the remaining data segment is smaller than the send segment size. The remaining data segment is sent to the receiver in response to identifying that the remaining data segment is a portion of the data block.

Smart nagling in a TCP connection

A dynamically-enforceable application-controlled quasi-reliable extension to TCP permits a client application to dynamically set a percent loss tolerance for data transmission reliability through network input/output system calls to the TCP, thereby programming the transport layer to optimistically acknowledge non-critical missing frames. The reliability requirement can be dynamically set within TCP to the level of reliability required for specific data frames within the data stream during the data transfer. Based on this loss tolerance specified, the TCP layer makes a determination whether to trigger a retransmission or continue delivering out-of-order frames to the application. A forced acknowledgement frame is sent for each missing packet until the number missing packets causing forced acknowledgments within the current receive buffer frame exceeds the loss tolerance. This process avoids needless retransmissions and permits the TCP data flow and sliding window to advance uninterrupted, thereby providing substantial performance benefits to network throughput.

Method, system and article for client application control of network transmission loss tolerance

In a computer system having memory, at least one processor and a physical input-output adapter, the at least one processor generates logical partitions, including local logical partitions and an input-output logical partition. The logical partitions have respective local virtual adapters. The input-output logical partition has a shared virtual adapter configured to communicate with the physical input-output adapter, such that a plurality of the local logical partitions share the physical input-output adapter via the shared virtual adapter. The at least one processor provides communication for ones of the local virtual adapters of the local logical partitions directly with the physical input-output adapter by a hypervisor.

Hypervisor-based data transfer

A system for consolidating TCP ports. In response to initiating a connection to a hidden port via a network, a TCP in TCP packet is created. Then, the TCP in TCP packet is sent to the hidden port via the network via a network visible port.

Method and system for consolidating TCP ports

TCP congestion avoidance is implemented upon retransmission of a packet and is reverted back to the original congestion state upon receipt of an early acknowledgement (ACK), indicating reordering of packets, thereby eliminating a needless restriction on TCP bandwidth. Upon receiving an ACK to a retransmitted packet, it is determined if the ACK resulted from receipt of the original reordered packet or the retransmitted packet, based on the arrival time of the ACK at the sender. If the round-trip-time (RTT) for the retransmitted packet is much lower than the average or current calculated RTT for the network link between sender and receiver, then the retransmission occurred as a result of a reordering event, and the congestion window is restored back to its value prior to the retransmission, thereby permitting the network link to continue operating at its original increased throughput.

Method, system and article for improved TCP performance during packet reordering

A method, apparatus, and computer instructions for processing packets in a network layer. A packet is received for transmission in which the packet includes a destination address. A routing entry for the destination address is identified. A determination is made as to whether a flag indicating the packet is to be broadcast has been set in the routing entry. Header information is placed in the packet in which the header information indicates that the packet is a broadcast packet in response to the flag being set.

Method and apparatus for processing outgoing internet protocol packets

A data processing system for resending a previously sent email message. A new recipient for the previously sent email message is selected. The new recipient of the previously sent email message is added in response to selecting the new recipient, wherein the previously sent email message is directed to the new recipient and a previous recipient. An indicator to the previously sent email message is added in response to selecting the new recipient, wherein the indicator indicates that the new recipient has been added to a form a new version of the previously sent email message. The new version of the previously sent email message is sent in response to adding the indicator, wherein the indicator causes the new version of the previously sent email message to replace the previously sent email message at an email program of the previous recipient.

Method and apparatus for adding recipients to sent email

A method, system, and program for monitoring thread usage to dynamically control a thread pool are provided. An application running on the server system invokes a listener thread on a listener socket for receiving client requests at the server system and passing the client requests to one of multiple threads waiting in a thread pool. Additionally, the application sends an ioctl call in blocking mode on the listener thread. A TCP layer within the server system detects the listener thread in blocking mode and monitors a thread count of at least one of a number of incoming requests waiting to be processed and a number of said plurality of threads remaining idle in the thread pool over a sample period. Once the TCP layer detects a thread usage event, the ioctl call is returned indicating the thread usage event with the thread count, such that a number of threads in the thread pool may be dynamically adjusted to handle the thread count.

Monitoring thread usage to dynamically control a thread pool

A method, system, and computer program product in a computer-readable medium for delivering data, received from a network, to a storage buffer assigned to an application is proposed. An application designates a communication buffer within a local data processing system for buffering data communicated with an application. The local data processing system reports to a network interface of the local data processing system a memory address of the designated communication buffer, and the data processing system creates a cookie containing the memory address. The data processing system then sends the cookie form the local data processing system to a remote data processing system, such that the remote data processing system may address data directly to the designated communication buffer.

Method, system and computer program product for delivering data to a storage buffer assigned to an application

A method for detecting and reducing packet drops in a communications system may include receiving a first packet having a sequence number, determining if a second packet having a sequence number immediately preceding the sequence number of the first packet is lost, and/or determining if a third packet having a sequence number immediately preceding the sequence number of the second packet is lost. A counter may be incremented if the second and third packets are determined as lost. An amount of time required before detecting link activity after an end of a previous frame may be reduced if the counter violates the count threshold and the amount of time has not been previously reduced. A lower link speed may be negotiated if a current link speed is above a negotiable threshold and the amount of time required before detecting link activity has been previously reduced.

Growing community of inventors

Austin, TX, United States of America

Kavitha Vittal Murthy Baratakke