Data management strategies for relative quality of service in virtualised storage systems

The amount of data managed by organisations continues to grow relentlessly. Driven by the high costs of maintaining multiple local storage systems, there is a well established trend towards storage consolidation using multi-tier Virtualised Storage Systems (VSSs). At the same time, storage infrastructures are increasingly subject to stringent Quality of Service (QoS) demands. Within a VSS, it is challenging to match desired QoS with delivered QoS, considering the latter can vary dramatically both across and within tiers. Manual efforts to achieve this match require extensive and ongoing human intervention. Automated efforts are based on workload analysis, which ignores the business importance of infrequently accessed data. This thesis presents our design, implementation and evaluation of data maintenance strategies in an enhanced version of the popular Linux Extended 3 Filesystem which features support for the elegant specification of QoS metadata while maintaining compatibility with stock kernels. Users and applications specify QoS requirements using a chmod-like interface. System administrators are provided with a character device kernel interface that allows for profiling of the QoS delivered by the underlying storage. We propose a novel score-based metric, together with associated visualisation resources, to evaluate the degree of QoS matching achieved by any given data layout. We also design and implement new inode and datablock allocation and migration strategies which exploit this metric in seeking to match the QoS attributes set by users and/or applications on files and directories with the QoS actually delivered by each of the filesystem’s block groups. To create realistic test filesystems we have included QoS metadata support in the Impressions benchmarking framework. The effectiveness of the resulting data layout in terms of QoS matching is evaluated using a special kernel module that is capable of inspecting detailed filesystem data on-thefly. We show that our implementations of the proposed inode and datablock allocation strategies are capable of dramatically improving data placement with respect to QoS requirements when compared to the default allocators.