You are developing a storage system for a application that demands unusually high reliability, so you have decided to use a three-replica durable storage scheme. You plan to use three ordinary disk drives D1, D2, and D3, and arrange that D2 and D3 store identical mirror copies of each block stored on D1. The disk drives are of a simple design that does not report read errors. That is, they just return data, whether or not it is valid.
10.1a. You initially construct the application so that it writes a block of data to the same sector on all three drives concurrently. After a power failure occurs during the middle of a write, you are unable to reconstruct the correct data for that sector. What is the problem?
10.1b. Describe a modification that solves this problem.
10.1c. One day there is a really awful power glitch that crashes all three disks in such a way that each disk corrupts one random track. Fortunately, the system wasn't writing any data at the time. Describe a procedure for reconstructing the data and explain any cases that your procedure cannot handle.