Senior Design Projects Disk Scrubbing

Signatures are a well-established technique to find differences between two different two storage containers. It was originally conceived to compare copies of two replicated databases. A similar technique is used in Tripwire to ascertain that system files are not changed.

This project combines both ideas, based on recent theoretical research by Witold Litwin and me. Assume a reasonably scaled storage system in which we provide redundancy by using Software RAID or Software p+q RAID. We maintain a signature map of superblocks on each disk. We can use the signatures in the signature map to check whether the contents of the disk are indeed what they are by going to the disk and reading the corresponding sectors and recalculating the signature.

The scope of this project is to verify whether the proposed algorithm is a performance bottleneck. For this purpose, the signature scheme will be implemented at a single (Linux) machine. Writes to disk will be intercepted and the signature map maintained. The storage performance will be evaluated. The results are to be published in a peer-reviewed journal or conference.

Limited funding is available. The project can be extended to provide a Master's Thesis for a member of the 5-year program.