Abstract: A system and method for tuning TCP/IP acknowledgments is provided. The system and method reduces the number of acknowledgments sent by a TCP/IP receiver by determining whether the connection state with the sender warrants using minimal acknowledgments. If minimal acknowledgments are used, the receiver sends fewer acknowledgments to the sender in response to received packets. The number of packets that are received before an acknowledgment is returned is increased until the delay value reaches a threshold value. The threshold value can be determined based on the size of the buffer setup to receive packets from the sender during the session. If errors, such as TCP/IP timeouts or duplicate packets, are detected, the threshold is changed to the last delay value that did not cause errors. If further errors are detected, the system is programmed to revert to sending traditional acknowledgments for the session.

Abstract: Network input processing is distributed to multiple CPUs on multiprocessor systems to improve network throughput and take advantage of MP scalability. Packets are received by the network adapter and are distributed to N receive buffer pools set up by the device driver, based on N CPUs being available for input processing of packets. Each receive buffer pool has an associated CPU. Packets are direct memory accessed to one of the N receive buffer pools by using a hashing function, which is based on the source MAC address, source IP address, or the packet's source and destination TCP port numbers, or all or a combination of the foregoing. The hashing mechanism ensures that the sequence of packets within a given communication session will be preserved. Distribution is effected by the network adapter, which sends an interrupt to the CPU corresponding to the receive buffer pool, subsequent to the packet being DMAed into the buffer pool.

Abstract: A method, system and apparatus for improving network performance of a multiprocessor system that has a limited number of physical network interfaces by using virtual IP addresses are provided. When a processor of the multiprocessor system determines that it is processing network data, it uses a virtual IP address to transmit the data. The virtual IP address is associated with a buffer; and thus the data is sent to the buffer. This allows the processor to be free to process other data, network or otherwise. The buffer, with the help of a controller, then contends for the use of one of the limited physical network interfaces to transmit the data.

Abstract: An apparatus and method for an improved bulk read socket call are provided. With the apparatus and method, a new field, so_rcvlen, is added to the socket structure that identifies the bulk read size requested by the user. The kernel of the prior art recv( ) function is also modified so that it sets the so_rcvlen to the size requested by the user prior to the recv( ) function going to sleep and waiting for the full data size requested by the user. A new flag, SP_MSGWAITALL, is also provided in the socket structure. In the TCP input processing, when data is received for a particular socket, the current setting of the SP_MSGWAITALL is checked. If the SP_MSGWAITALL flag is set, it is determined whether the amount of data stored in the socket receive buffer is less than the value of so rcvlen. If not, the TCP input processing does not wake up the recv( ) thread. However, for every alternate segment, the TCP input processing sends back an acknowledgment (ACK).

Abstract: Network input processing is distributed to multiple CPUs on multiprocessor systems to improve network throughput and take advantage of MP scalability. Packets received on the network are distributed to N high priority threads, wherein N is the number of CPUs on the system. N queues are provided to which the incoming packets are distributed. When one of the queues is started, one of the threads is scheduled to process packets on this queue at any one of the CPUs that is availableat the time. When all of the packets on the queue are processed, the thread becomes dormant. Packets are distributed to one of the N queues by using a hashing function based on the source MAC address, source IP address, or the packet's source and destination TCP port number, or all or a combination of the foregoing. The hashing mechanism ensures that the sequence of packets within a given communication session will be preserved. Distribution is effected by the device drivers of the system.