Implementation and evaluation of consistent online backup in transactional file systems

A consistent backup which preserves data integrity across files in a file system is of utmost importance for the purpose of correctness and minimizing system downtime during the process of data recovery. With present day demand for continuous access to data, backup has to be taken of an active file system, putting the consistency of the backup copy at risk. In order to address this issue, we propose a scheme which is referred to as mutual serializability assuming that the file system support transactions. Mutual serializability captures a consistent backup of an active file system by ensuring that the backup transaction is mutually serializable with every other transaction individually. This mutually serializable relationship is established considering an extended set of conflicting operations which include read-read conflicts. User transactions serialize within themselves using some standard concurrency control protocol such as the Strict 2PL and a set of conflicting operations that only include the traditional read-write, write-write and write-read conflicts. The proposed scheme has been implemented on top of a transactional file system and workloads exhibiting a wide range of access patterns were used as inputs to conduct experiments in two scenarios, one with the mutual serializability protocol enabled (thus capturing a consistent backup) and one without (thus capturing an inconsistent backup). The results obtained from the two scenarios were then compared to determine the overhead incurred while capturing a consistent backup. The performance evaluation shows that for workloads resembling most present day real workloads exhibiting low intertransactional sharing and actively accessing only a small percentage of the entire file system space, the proposed scheme has very little overhead (a 5.7% increase in backup time and a user transaction throughput reduction of 3.68%). Noticeable performance improvement is recorded when using performance enhancing heuristics which involve diversion of the backup transaction to currently “colder” regions of the file system hierarchy on detecting conflicts with user transactions.