Fault tolerance in a parallel database system

A computer program product for implementing a method for detecting the health of parallel database services, the computer program product comprising one or more computer-readable storage media having stored thereon computer-executable instructions that, when executed by one or more processors of the computing system, cause the computing system to perform the method, the method comprising:

• an act of sending an inquiry to one or more embedded health detectors embedded in a specified parallel database service to determine the health of the parallel database service;
• an act of receiving from the one or more embedded health detectors a service status indication for the specified service;
• based on the received service status indication, an act of determining that the service is to be at least one of restarted and terminated;
• and
• based on the determination, an act of sending at least one of a restart and a terminate command to the specified service.

A computer system comprising the following:

• one or more processors;
• system memory;
• one or more computer-readable storage media having stored thereon computer-executable instructions that, when executed by the one or more processors, causes the computing system to perform a method for establishing a fault tolerant parallel database system, the method comprising the following:
• an act of establishing a control node cluster comprising at least one active control node and at least one spare control node, wherein each node of the control node cluster includes specific functions assumable only by other control nodes;
• an act of establishing a compute node cluster comprising at least one active computing node, at least one spare computing node, at least one active storage node and at least one spare storage node, wherein each of the computing and storage nodes includes specific functions assumable only by other computing and storage nodes, and wherein the control node cluster and the compute node cluster are located on the same computing appliance;
• an act of detecting a failure of at least one active node;
• and
• an act of instantiating a corresponding spare node that is configured to perform the functions of the failed active node, wherein one or more service policies are implemented at the computing appliance which define how failures are to be handled for each type of component on the computing appliance.