Research On I/O Optimization Technology Based On Solid State Disk

Information technologies have been applied to many aspects of our everyday life. Human society can move ahead continuously with the support of various information services, applications and massive data. This promotes the rapid growth of data capacity and types, and improvement of computer organization, scale and performance. Many fields such as scientific research, social life and so on will produce huge amounts of new data every day. How to storage the data resource fast, efficiently and safely has become one of the research hot spot and the biggest challenge.I/O performance of the storage system has long been the bottleneck of computer system composed with traditional mechanical disk HDD. In recent years, flash based solid state disk (SSD) technology has been greatly developed. Because it is composed completely by the semiconductor chips, no mechanical components, so it has high performance, high reliability, random access, low power consumption etc. And it is expected to eliminate the performance bottleneck of storage system so to bring fundamental changes to the storage system. Although SSD has obvious performance advantages, it has different characteristics and disadvantages for its internal structure. With the semiconductor technology more and more mature and SSD are applied in the storage system more extensively and importantly, study how to maximize the use of SSD in modern storage systems, while avoiding their inherent defects, improve the performance and reliability of storage systems, and reduce the energy consumption has the important research significance.This paper studies how to apply SSD in modern storage systems from the following aspects, to improve the reliability and performance, while reducing the energy consumption of the storage system.Instead of only either SLC or MLC, a novel internally hybrid SSD architecture (SMARC) which consists of both SLC and MLC flash memory packages is proposed. By periodically migrating data between the two different portions (SLC and MLC space) according to running workloads dynamics, the proposed architecture is able to exploit their respective advantages complementarily to help overall performance, reliability and energy consumption. Our simulation results of a variety of workloads have shown that SMARC can improve the performance, at the same time, significantly enhancing reliability and reducing energy consumption. Furthermore, SMARC can be easily extended to provide differentiated storage services if that is desired, which bears important implications for modern storage systems.A new SSD I/O scheduler which takes SSD's peculiarities into account in its design is developed. In more detail, firstly, we partition the entire SSD space into several regions as basic scheduling units and issue requests to them simultaneously to exploit internally rich parallelism. Secondly, we give preference to read requests over writes to take advantage of the fact that reads are much faster than writes, avoiding excessive read-blocked by-write interference. Lastly, within each region's scheduling queue, we sort write requests before issuing them in anticipation of transforming random writes to sequential patterns to avoid harmful random writes. Our evaluation results of a variety of workloads have demonstrated that Regional Scheduler is capable of improving the performance by 17-32% compared with the best of the four kernel I/O schedulers, in the meantime, improving SSD's lifetime because of the successful transformation of random writes to sequential writes.Flash Saver, which is coupled with the ext2/3 file system and aims to significantly reduce the write traffic to SSDs, is presented. Flash Saver deploys deduplication and delta-encoding to reduce the write traffic. Specifically, Flash Saver applies deduplication to file system data blocks and delta-encoding to file system meta-data blocks, based on two important observations which are:(1) There exist large amounts of duplicate data blocks; (2) Metadata blocks are accessed/modified much more frequently than data blocks, but with very minimal changes for each update. Experimental results have shown that Flash Saver can save up to 63% of the total write traffic, which implies reasonably prolonged lifetime, larger effective flash space and higher reliability than that of the original counterpart within their allowable lifespan.REST, an energy-efficient cloud storage system which is built upon a cluster-based object store similar to GFS is proposed. It achieves high energy-efficiency by cleverly exploiting the redundancy already present in the system without compromising the inherently well-established schemes for consistency, fault-tolerance, reliability, availability, etc, while maintaining a reasonable performance level. By modifying slightly the data-layout policy, REST can safely keep a large amount of the storage nodes in standby mode or even powered off entirely most of the time. Deploying a sophisticated monitor, it also provides the flexibility to power up sleeping or powered down nodes when necessary to accommodate to the variations in workloads. Trade-offs between energy efficiency and performance can be conveniently made by simply adjusting a trade-off metric in REST.Through the above four aspects, aimed at the main defects in the solid state disk system, research how to maximize use of SSD, improve the performance and reliability of storage system in modern storage systems from.This thesis presents comprehensive researches on the I/O optimization technology based on solid state disk from the aspects of changing the data layout strategy and storage system structure, reduce writing flow to SSD, design the I/O scheduling strategy of SSD disk inside, and combine the solid state disk and mechanical disk. The experimental results show that these methods can optimize the I/O performance of the SSD, make the best use of SSD, improve the storage system performance and reliability, while reducing energy consumption.