University at Buffalo
The State University of  New York
CCR
Center for Computational Research
Load Balancing
Fault Tolerance
n Fault Tolerance
q A platform has gone offline or lost network connectivity
m The grid-enabled SnB implementation is extremely fault tolerant
m The requested SnB trial results will be completed automatically even if all but one platform fail
¢ One node server has the ability to:
¢ Complete it’s platform Block Queue
¢ Process the Float Queue
¢ Process all other failed platform Block Queues
¢ Process failed trials from it’s own platform Block Queue
¢ Automatically !
m