This paper presents a method, infrastructure, and prototype that enables adaptive application task migration among a Grid environment. Our infrastructure automatically recon.gures distributed applications in response to network performance failures and denial of service (DoS) attacks. Through the use of performance monitoring software we enable network connection failover and automatic application task migration within a heterogeneous distributed computing environment. Our system distinguishes itself from other available adaptive computing frameworks because it is wholly composed of open source Grid-enabled components capable of both transparent and dynamic selection of message passing transports based on resource performance.