Predictive Prefetching for Parallel Hybrid Storage Systems

Abstract

In this paper, we present a predictive prefetching mechanism that is based on probability graph approach to perform prefetching between different levels in a parallel hybrid storage system. The fundamental concept of our approach is to invoke parallel hybrid storage system's parallelism and prefetch data among multiple storage levels (e.g. solid state disks, and hard disk drives) in parallel with the application's on-demand I/O reading requests. In this study, we show that a predictive prefetching across multiple storage levels is an efficient technique for placing near future needed data blocks in the uppermost levels near the application. Our PPHSS approach extends previous ideas of predictive prefetching in two ways: (1) our approach reduces applications' execution elapsed time by keeping data blocks that are predicted to be accessed in the near future cached in the uppermost level; (2) we propose a parallel data fetching scheme in which multiple fetching mechanisms (i.e. predictive prefetching and application's on-demand data requests) can work in parallel; where the first one fetches data blocks among the different levels of the hybrid storage systems (i.e. low-level (slow) to high-level (fast) storage devices) and the other one fetches the data from the storage system to the application. Our PPHSS strategy integrated with the predictive prefetching mechanism significantly reduces overall I/O access time in a hybrid storage system. Finally, we developed a simulator to evaluate the performance of the proposed predictive prefetching scheme in the context of hybrid storage systems. Our results show that our PPHSS can improve system performance by 4% across real-world I/O traces without the need of using large size caches.

Performance Optimization for Distributed Hybrid Storage Systems using a Predictive Approach

“…As we have the small cache implemented in the SS level (SS Cache), our data prefetch mechanism will cache copies of prefetched data beside the application's data read request the are read from the HD (SS misses). No moves for the entire data blocks from the HD to the SS takes place as that will waste the SS level's capacity and we will consume the parallel storage' bandwidth; especially if some of these data blocks need to be returned to the HD level later [2]. Using our SSs and HDs installed in our research lab's distributed storage setting, we found that SS provides higher performance than HD when using big sized data blocks.…”

“…Predictive approach predicts the data that will be requested in the future by the application based on its historical requests' patterns. An approach used to record the history of the application's previous data requests is called probability graph (PG) [2][6] [13]. Markov chain prediction approach was also used by other researchers [8].…”

“…It also performs LRU policy when the SS cache reaches to its full capacity. The predictor receives each application data read request and executes the probability graph-predictive algorithm to determine the data blocks that have good cache to be requested by the application soon in the near future [2]. Those requests are sent to the disk manager to read them from the HD to the SS unless if the data block is already existing in the SS level.…”

“…Mainly, there are two prefetch methods; predictive and informed. In predictive method, the system predicts what data will the application needs soon based on the patterns of previous requests history [2]. On the other hand, informed method makes use of the hints that are given from the application to fetch the data in advance [1][3] [5].…”

“…On the other hand, informed method makes use of the hints that are given from the application to fetch the data in advance [1][3] [5]. This research implements probability graph predictive method (PG) that continuously record the history of the data requested by the application at the client-side node [2] [6]. We study the performance impact of this scheme on distributed hybrid (multi-level) storage environment.…”