The Research And Implementation Of Checkpoint Technology Based On WinNT

A Checkpoint facility enables the intermediate state of a running process to be saved to stable storage. When a failure occurs, Users can resume execution of the process from the checkpoint file. This prevents the loss of data generated by long-running processes due to program or system failures.With the WinNT used widely, there are more and more applications based on it. People attach much importance to fault-tolerance of the Applications. As WinNT is not an open OS, it is hard to modify its kernel. So many research organizations have developed user-level checkpoint systems to improve the fault-tolerance of the application on it.At the beginning, this paper generally reviews the checkpoint technology, then analyzes and points out the limitation about the existing checkpoint system, then concludes the difficulty and the problems need to resolve. Secondly, this thesis introduces the key technology of checkpoint system, such as injecting DLL, intercepting and wrapping API, checkpointing and resuming the state of the opened file etc.Following, this paper makes a detailed analysis of the design and implement of the checkpoint system based on WinNT named NTckpt. It is composed of two modules. One is InjectDLL the other is T-FAULT.DLL. InjectDLL module can inject a DLL into an exe file transparently.T-FAULT.DLL is the kernel of NTckpt. It implements intercepting and wrapping API, saving and resuming the state of process etc.T-FAULT.DLL introduces a way to restore address space of process consistently. Using this way can avoid hidden trouble of the existing method but also simplify the resume of the process.At last, this paper adopts a method using Time Series Analysis to model and forecast the state size of process, which helps to select appropriate interval of checkpoint dynamically and reduces the overhead.