Recent increases in CPU performance have surpassed those in hard drives. As a result, disk operations have become more expensive in terms of the number of CPU cycles spent waiting for them to complete. File prediction can mitigate this problem by prefetching files into cache before they are accessed. Identifying relationships between individual files plays a key role in successfully performing file prefetching. It is well-known that previous patterns of file references can be used to predict future references. Nevertheless, knowledge about the programs producing the relationships between individual files has rarely been investigated. We present a Program-Based Successor (PBS) model that identifies relationships between files through the names of the programs accessing them. We develop a Program-based Last Successor (PLS) model derived from PBS to do file prediction. Our simulation results show that PLS makes 21 % fewer incorrect predictions and roughly the same number of correct predictions as the Last-Successor (LS) model. We also examine the cache hit ratio achieved by applying PLS to the Least Recently Used (LRU) caching algorithm and show that a cache using PLS and LRU together can perform better than a cache up to 40 times larger using LRU alone. Finally, we argue that because program-based successors are more likely to be used soon, incorrectly prefetched program-based successors are more likely to be used and thus less incorrect than incorrectly prefetched files from non-program-based models.

Cloud computing has become more and more popular nowadays. Both governments and enterprises provide service through the construction of public and private clouds accordingly. Among the platforms used in cloud computing, Hadoop is considered one of the most practical and stable systems. Nevertheless, as with other regular software, Hadoop still needs to rely on the underlying operating system to communicate with hardware to function appropriately. For modern computer systems, CPUs excessively outrun hard drives (hard disks). The computer hard disk has become a major bottleneck to the overall system performance. Consequently, computer programs can execute faster if their corresponding I/O operation can be completed sooner. This is important in particular when we want to expedite the execution of urgent programs in a busy system. Unfortunately, under the current Hadoop environment, users cannot prioritize operation of disk and memory for programs which they would like them to run faster. With the help of prioritized I/O service we developed earlier, we proposed and implemented a Hadoop environment with the ability of providing prioritized I/O service. Our Hadoop environment could accelerate the execution of programs with high priority assigned by users. We evaluated our design by executing prioritized programs in environments with different busy levels. Experimental results show that programs can improve their performance by up to 33.79% if executed with high priority.

SummaryComputers are indispensable to modern human society. Often, computers host multiple programs running simultaneously. However, among those programs, some maybe more time critical than others to users. Consequently, users would hope those time‐critical programs to finish their execution as soon as possible. Generally speaking, the course of program execution includes cooperative operation of CPU, memory, and hard disk. For CPU operation, modern computer systems have the ability to adjust the CPU scheduling sequence according to program priority. Nevertheless, most of them do not have effective ways to provide memory management and disk operation based on program priority. Consequently, it is hard to speed up the execution of prioritized programs as users would expect. The Linux operating system has been widely used in recent years. Even though it still does not support fully‐prioritized memory management and disk operation. Previously, based on the default Linux disk scheduler, complete fair queuing (CFQ), we proposed a new disk scheduling algorithm called prioritized CFQ (PCFQ) to provide fully‐prioritized disk operation. PCFQ had demonstrated good performance improvement for programs with high priority. In this paper, we report our effort to integrate PCFQ with prioritized memory management as a whole in the Linux operating system. With the cooperation between prioritized memory management and prioritized disk operation, our system further enhanced the performance that PCFQ achieved by up to 62.08%. Experimental results also showed that, compared with the current Linux system, our design can lift the performance of individual prioritized programs by up to 76.63% when there was one prioritized program on the system. The number went up to 82.45% when the system hosted three programs with high priority.Concurrency and Computation: Practice and Experience.Copyright © 2014 John Wiley & Sons, Ltd.

No abstract

No abstract