RFS: Repairable File Service

Faculty: Tzi-cker Chiueh

Members:

Project Description

The data contents of an information system may be corrupted due to security breaches or human errors. The financial loss of such corruption is typically proportional to the amount of time required to recover the system's data/service. RFS is specifically designed to facilitate the reparation of compromised network file servers. An architectural innovation of RFS is that it is decoupled from and requires no modifications on the shared file server that is being protected. Repairable File Service (RFS) is not a file system on its own. It is designed to be a general framework for protecting shared file servers from irrevocable damages caused by errors or attacks. In the normal operating mode, RFS maintains a file update log and an inter-process dependency log. In the repair mode, RFS first determines the exact extent of system damage, and then performs selective roll-back of those data blocks that are considered contaminated. Although the design of RFS centers around the same two fundamental tasks associated with any repairable information system, it is also driven by a set of design goals:

RFS should operate in a way that does not require any modifications to the shared file server that it is protecting,
RFS should not introduce significant performance overhead to the file access path that disrupts the interactions between the shared file server and its clients, and
The system architecture of RFS should be sufficiently modular that the components which are independent of the underlying network file acces protocol (e.g., NFS) should be reusable across different network file access protocols.

System Architecure

Traffic Interceptor logs all file update commands sent to the protected shared file server, Versioning File System records all file system updates, Analysis and Repair Engine analyze the data updating history and operation dependencies, Syscall Logger is embedded in the NFS clients,

Current Work

Optimization of the cleaning algorithm of Versioning File System

File System Evaluation

Trace based evaluation
NFS traces from Harvard
SPEC SFS 3.0(SFS97_R1)

Publications

Ningning Zhu, Jiawu Chen, Tzi-cker Chiueh. ``TBBT: Scalable and Accurate Trace Replay for File Server Evaluation,'' in Proceedings of USENIX Conference on File and Storage Technology (FAST 2005), San Francisco, CA., December 2005.
Ningning Zhu and Tzi-cker Chiueh, "Design, Implementation and Evaluation of Repairable File Service" , The International Conference on Dependable Systems and Networks,San Francisco, CA, June 22nd - 25th, 2003.

Technical Report

Ningning Zhu and Tzi-cker Chiueh, "Efficient and Portable User-Level File System Update Logging", Submitted to The International Conference on Dependable Systems and Networks (2005),

Tzi-cker Chiueh, Ningning Zhu and Dhruv Pilania, How to Repair Compromised Information Systems Quicky?, submitted to HotOS IV.
Research Proficiency Exam Report, survey of Data Versioning Systems