Importance

Loading Importance Indicators...

Abstract

A clustered computer system includes multiple computer systems (or nodes) on a network that can become members of a group to work on a particular task, referred to herein as a protocol. A protocol is defined so that each phase of the protocol is terminated with an acknowledge (ACK) round. Within each phase of the protocol, a node cannot both send and receive a data message. The protocol includes an ACK round that provides a relative time event that indicates when a data message should be received. If the data message is not received when the ACK round occurs, the receiver knows that the sender did not send it, and can request that the sender re-send the missing data message. In a first embodiment, referred to herein as the 'post-ACK' case, the receipt of expected data messages is checked after the ACK round occurs. In a second embodiment, referred to herein as the 'pre-ACK' case, the receipt of expected data messages is checked before the ACK round occurs. In both cases, when the receiver sees the ACK round, it knows that the sender sent the data message. If the data message was not received by the receiver, it knows to request that the sender re-send the data message. In this manner the preferred embodiments provide an architected way for sending and receiving data messages without using timers in a clustered computing environment that includes ordered messages. As a result, the preferred embodiments may be readily implemented on computer clusters that include nodes on a wide area network (WAN) that have a large variability in their response times.

Loading the Abstract Image...

First Claim