Towards Workload-Driven Adaptation of Data Organization in Heterogeneous Storage Systems
Collecting, managing, and analyzing huge data sets (Big Data) in science and industry poses challenges to current data storage systems in terms of storage capacity, performance, and reliability. In particular, the I/O performance may be a key factor to speed up the data analysis. However, the performance of a storage system significantly depends on its configuration and the access pattern. Designing storage systems always implies making compromises between performance, fault tolerance and net capacity. The decision which compromise is made (which RAID level is used) has to be taken at deployment time because runtime reconfigurations are usually prohibitively expensive (due to coarse granularity) in current storage architectures.
In this paper, we propose a workload-driven approach to adaptive reconfiguration covering the functionality of the file system, volume manager and RAID. Our approach enables fine-grained reconfigurations of the data organization of files and file fragments to adapt the storage system to changing workloads, while considering the different characteristics of the storage devices ( SSDs and HDDs) in a heterogeneous storage system. We first discuss how our approach decreases the costs of adaptations compared to existing approaches making a continuous and effective adaptation feasible, even for large volumes of data. Then, we present an evaluation based on a prototypical implementation confirming the benefits of our approach.
KeywordsStorage System Access Pattern Data Organization Spare Capacity Data Layout
Unable to display preview. Download preview PDF.
- 1.Appuswamy, R., van Moolenbroek, D.C., Tanenbaum, A.S.: Block-level RAID is dead. In: Proceedings of the 2nd USENIX Conference on Hot Topics in Storage and File Systems, HotStorage 2010, pp. 1–5. USENIX Association, Berkeley (2010)Google Scholar
- 2.Appuswamy, R., van Moolenbroek, D.C., Tanenbaum, A.S.: Integrating flash-based SSDs into the storage stack. In: Proceedings of the IEEE 28th Symposium on Mass Storage Systems and Technologies (MSST 2012), pp. 1–12. IEEE (2012)Google Scholar
- 3.Axboe, J.: Flexible I/O tester (2013), http://freecode.com/projects/fio
- 4.Guerra, J., Pucha, H., Glider, J., Belluomini, W., Rangaswami, R.: Cost effective storage using extent based dynamic tiering. In: Proceedings of the 9th USENIX Conference on File and Storage Technologies, FAST 2011, pp. 1–14. USENIX Association, Berkeley (2011)Google Scholar
- 5.Jacob, B.L., Ng, S.W., Wang, D.T.: Memory Systems: Cache, DRAM, Disk. Morgan Kaufmann (2008)Google Scholar
- 7.Jeremic, N., Mühl, G., Busse, A., Richling, J.: The pitfalls of deploying solid-state drive RAIDs. In: Proceedings of the 4th Annual International Conference on Systems and Storage (SYSTOR 2011), pp. 14:1–14:13. ACM (June 2011), http://doi.acm.org/10.1145/1987816.1987835
- 10.Seagate Technology: Product manual Constellation. 2 SAS. Tech. rep., Seagate Technology (April 2012), http://www.seagate.com/files/www-content/support-content/documentation/product-manuals/en-us/Enterprise/Constellation_2_5in/100620418h.pdf