Fast transparent fault shift device and implementing method facing to service

Abstract

Based on character of general-purpose operating system with kernel state and user state, the invention establishes device for transferring fault. The said device includes main control process module (MCPM) and comm control process module (CCPM) in operation system (OP) of main and standard servers. Running in user state of OP, MCPM is in charge for controlling state switching of service software as well as controlling transfer sequence and data synchronization of other application processes (including comm control process). Running in kernel state of OP, CCPM is in charge for switching connection of bottom network, and processing message buffered in comm link. The invention guarantees that when fault occurs, fault transfer carried by system is transparent to user. Possessing short time delay of fault transfer and lower system overhead, the invention carries out fault transfer in high efficiency and transparency under precondition of ensuring service usability.

Description

technical field [0001] The invention relates to the field of server failover, in particular to a service-oriented fast and transparent failover device and an implementation method. Background technique [0002] With the continuous development of network communication technology, various network value-added services and application services are also growing rapidly. Users not only need open and complete service provision, but also high-quality and uninterrupted service delivery. This puts forward strict requirements for fault-tolerant technology in a service-oriented environment. In key network services, if system equipment fails or is maintained offline, the service will be interrupted, resulting in serious consequences. [0003] A well-known example is e-bay. In 1999, e-bay's 10-hour network service interruption caused at least five million dollars in economic losses. It can be seen that in a service-oriented environment, critical network services must be able to guarantee...

AI Technical Summary

Problems solved by technology

[0006] At present, there are not many researches on failover, and most failover schemes cannot guarantee the availability of services, and will be more or less It causes less service interruption and is not transparent to users. Clients are often involved in the failover and recovery process. The client's participation may be a simple timeout re-request, or it may directly participate in server selection and connection state migration. Although A solution with client participation will provide some flexibility, but a solution that is transparent to the client is more necessary

[0007]At the same time, the data transmission and information interaction between the primary and secondary servers are usually completed using TCP (Transmission Control Protocol-transport layer control protocol), the TCP protocol processing is complex and inefficient High, will increase the failover delay and system overhead

[0008] Finally, since most operating systems do not provide a failover mechanism internally, the failover scheme is basically implemented in the application program, which leads to High system processing overhead

Images