Xiao Qin's Research

Auburn University

Final Report

BUD: A Buffer-Disk Architecture for Energy Conservation in Parallel Disk Systems

Download BUD Final Report PDF




Our Mission: Parallel disks consisting of multiple disks with high-speed switched interconnect are ideal for data-intensive applications running in high-performance computing systems. Improving the energy efficiency of parallel disks is an intrinsic requirement of next generation high-performance computing systems, because a storage subsystem can represent 27% of the energy consumed in a data center. However, it is a major challenge to conserve energy for parallel disks and energy efficiently coordinate I/Os of hundreds or thousands of concurrent disk devices to meet high-performance and energy-saving requirements. The goal of this research is to develop energy conservation techniques that will provide significant energy savings while achieving low-cost and high-performance for parallel disks.


Research and Educational Activities


Project team meetings held

The PI and Co-PI met together on several occasions during the past year: March 30, 2007, at Long Beach, California; and August 6, 2007, Arlington VA (during the NSF FSIO workshop). These project meetings are playing an important role in coordinating our collaborative research activities. In addition, the PI, Co-PI, and senior personnel made use of frequent emails and phone discussions to collaborate on the project.

A buffer-disk architecture

We designed an energy-efficient buffer-disk architecture called BUD for parallel disk systems. In this BUD configuration, referred to as interBUD aims to support inter-request parallelisms.

Power consumption models

We started the power consumption modeling effort by introducing the simplest model where each disk has only two power states: a working state Sw and a sleep state Ss. All the disks in a parallel disk system will be modeled by their power characteristics, i.e., D = (pw, ps, pu, pd, tu, td), where pw is the power consumed in the working state, ps is the power consumed in the working state, pu is the power consumed during wake up, pd is the power consumed during shutdown, tu is the transition time from the sleep state to working state. td is the transition time from the working state to sleep state. The total energy consumed by a disk can be calculated as , where Ew and Es are the energy consumed by the disk when it is in the working sleep states, Eu and Ed is the energy consumed when the disk transits from the sleep to working state and vice versa.

Energy-aware prefetching (PRE-BUD)

Our goal in this part of study was to use the interBUD architecture to aid in the reduction of the power consumption for these large-scale parallel disk systems. This work aimed at reducing the energy consumed of these systems by comparing two different approaches at reducing energy consumption using a disk management scheme called PRE-BUD. PRE-BUD is able to prefetch data into buffer disks with a desired consequence of reducing the total energy consumption of a large-scale parallel disk system. The first approach using PRE-BUD examined adds an extra disk, which becomes the buffer disk, and the other approach uses an existing disk as the buffer disk. This algorithm relies on the fact that in some applications a small percentage of the data is frequently accessed. Our goal was to place a small amount of frequently accessed data into the buffer disk reducing energy consumption. This work also focused on increasing the reliability of the disk system over typical energy aware disk management schemes. This can be achieved with a reduction in state transitions that the use of a buffer disk facilitates.

Write request processing

In the BUD architecture, large and small writes are processed in different ways. Large write requests are issued directly to data disks. In contrast, small write requests are sent to an active buffer disk. Once the data of a write request is transferred to buffer or data disks, an acknowledgement is returned to an application issuing the request. Our study confirmed that seek times of small disk request dominates disk I/O processing times. To alleviate this situation, we made use of sequential accesses, i.e., a log file system, in interBUD, thereby making the seek time of most write requests to be zero. This is because disk head of a sequential access disk is, in most cases, positioned on an empty track that is available for incoming write requests. The seek times of write requests handled by buffer disks are zero unless the buffer disks are in a process of moving data to data disks or responding read requests.

Energy-efficient data placement

We investigated data placement strategies, which place all data sets onto a disk array before they are accessed. Data placement is one of avenues that can significantly affect the overall performance of a parallel I/O system. We developed a data placement scheme that can simultaneously achieve high energy efficiency and quick response times through intelligent in our BUD architecture.

A Prefetching Scheme for Energy Conservation in Parallel Disk Systems

Large-scale parallel disk systems are frequently used to meet the demands of information systems requiring high storage capacities. A critical problem with these large-scale parallel disk systems is the fact that disks consume a significant amount of energy. To design economically attractive and environmentally friendly parallel disk systems, we developed two energy-aware prefetching strategies for parallel disk systems with disk buffers. First, we introduced a new buffer disk architecture that can provide significant energy savings for parallel disk systems while achieving high performance. Second, we designed a prefetching approach to utilize an extra disk to accommodate prefetched data sets that are frequently accessed. Third, we developed a second prefetching strategy that makes use of an existing disk in the parallel disk system as a buffer disk. Compared with the first prefetching scheme, the second approach lowers the capacity of the parallel disk system. However, the second approach is more cost-effective and energy-efficient than the first prefetching technique. Finally, we quantitatively compare both of our prefetching approaches against two conventional strategies including a dynamic power management technique and a non-energy-aware scheme. Using empirical results we show that our novel prefetching approaches are able to reduce energy dissipation in parallel disk systems by 44% and 50% when compared against a non-energy aware approach. Similarly, our strategies are capable of conserving 22% and 30% of the energy when compared to the dynamic power management technique.

Improving Reliability and Energy Efficiency of Disk Systems via Utilization Control

As disk drives become increasingly sophisticated and processing power increases, one of the most critical issues of designing modern disk systems is data reliability. Although numerous energy saving techniques are available for disk systems, most of energy conservation techniques are not effective in reliability critical environments due to their limitation of ignoring the reliability issue. A wide range of factors affect the reliability of disk systems; the most important factors – disk utilization and ages – are the focus of this study. We build a model to quantify the relationship among the disk age, utilization, and failure probabilities. Observing that the reliability of a disk heavily relies on both disk utilization and age, we propose a novel concept of safe utilization zone, where energy of the disk can be conserved without degrading reliability. We investigate an approach to improving both reliability and energy efficiency of disk systems via utilization control, where disk drives are operated in safe utilization zones to minimize the probability of disk failure. In this study, we integrate an existing energy consumption technique that operates the disks at different power modes with our proposed reliability approach. Experimental results show that our approach can significantly improve reliable while achieving high energy efficiency for disk systems.

DARAW: A New Write Buffer to Improve Parallel I/O Energy-Efficiency

In the past decades, parallel I/O systems have been used widely to support scientific and commercial applications. New data centers today employ huge quantities of I/O systems, which consume a large amount of energy. Most large-scale I/O systems have an array of hard disks working in parallel to meet performance requirements. Traditional energy conservation techniques attempt to place disks into low-power states when possible. In this part of research, we proposed a novel strategy, which aims to significantly conserve energy while reducing average I/O response times. This goal was achieved by making use of buffer disks in parallel I/O systems to accumulate small writes to form a log, which can be transferred to data disks in a batch way. We develop an algorithm - dynamic request allocation algorithm for writes or DARAW - to energy efficiently allocate and schedule write requests in a parallel I/O system. DARAW is able to improve parallel I/O energy efficiency by the virtue of leveraging buffer disks to serve a majority of incoming write requests, thereby keeping data disks in low-power state for longer period times. Buffered requests are then written to data disks at a pre-determined time. Experimental results show that DARAW can significantly reduce energy dissipation in parallel I/O systems without adverse impacts on I/O performance.

Heat-Based Dynamic Data Caching: A Load Balancing Strategy for Energy-Efficient Parallel Storage Systems with Buffer Disks

Performance improvement and energy conservation are two conflicting objectives in large scale parallel storage systems. In this project, we proposed a novel solution to achieve the twin objectives of maximizing performance and minimizing energy consumption of parallel storage systems. Specifically, a buffer-disk based architecture (BUD for short) is designed to conserve energy. A heat-based dynamic data caching strategy is developed to improve performance. The BUD architecture strives to allocate as many requests as possible to buffer disks, thereby keeping a large number of idle data disks in low-power states. This can provide significant opportunities for energy conservation while making buffer disks a potential performance bottleneck. The heat-based data caching strategy aims to achieve good load balancing in buffer disks and alleviate overall performance degradation due to unbalanced workload. Our experimental results show that the proposed BUD framework and dynamic data caching strategy are able to conserve energy by 84.4% for small reads and 78.8% for large reads with slightly degraded response time.

An Adaptive Energy-Conserving Strategy for Parallel Disk Systems

In the past decade parallel disk systems have been highly scalable and able to alleviate the problem of disk I/O bottleneck, thereby being widely used to support a wide range of data- intensive applications. Optimizing energy consumption in parallel disk systems has strong impacts on the cost of backup power-generation and cooling equipment, because a significant fraction of the operation cost of data centres is due to energy consumption and cooling. Although a variety of parallel disk systems were developed to achieve high performance and energy efficiency, most existing parallel disk systems lack an adaptive way to conserve energy in dynamically changing workload conditions. To solve this problem, we develop an adaptive energy-conserving algorithm, or DCAPS, for parallel disk systems using the dynamic voltage scaling technique that dynamically choose the most appropriate voltage supplies for parallel disks while guaranteeing specified performance (i.e., desired response times) for disk requests. We conduct extensive experiments to quantitatively evaluate the performance of the proposed energy-conserving strategy. Experimental results consistently show that DCAPS significantly reduces energy consumption of parallel disk systems in a dynamic environment over the same disk systems without using the DCAPS strategy.

Energy-Efficient Data Placement

We investigated data placement strategies, which place all data sets onto a disk array before they are accessed. Data placement is one of avenues that can significantly affect the overall performance of a parallel I/O system. First, we developed a static non-partitioned file assignment strategy for parallel I/O systems, called static round-robin (SOR), which is immune to the workload assumption. Next, to achieve energy-conservation and prompt responses simultaneously, we designed an energy-aware strategy, called striping-based energy-aware (SEA), which can be integrated into data placement in RAID-structured storage systems to noticeably save energy while providing quick responses. Finally, to illustrate the effectiveness of SEA, we implemented two SEA-powered striping-based data placement algorithms, SEA0 and SEA5, by incorporating the SEA strategy into RAID-0 and RAID-5, respectively.

An Energy-Aware Data Reconstruction Strategy for Mobile Disk Arrays

Compared with conventional stationary storage systems, mobile disk-array-based storage systems are more prone to disk failures due to their severe application environments. Further, they have very limited power supply. Therefore, data reconstruction algorithms, which are executed in the presence of disk failure, for mobile storage systems must be performance-driven, reliability-aware, and energy-efficient. We developed a reconstruction strategy, called multi-level caching-based reconstruction optimization (MICRO), which can be applied to RAID-structured mobile storage systems to noticeably shorten reconstruction times and user response times while saving energy.

Understanding the Relationship between Energy Conservation and Reliability in Parallel Disk Arrays

Power management and workload skew based energy conservation schemes for disk arrays inherently and adversely affect the reliability of disks due to either workload concentration or frequent disk speed transitions. A thorough understanding of the relationship between energy saving techniques and disk reliability is indispensible. We developed an empirical reliability model, called Predictor of Reliability for Energy Saving Schemes (PRESS). Fed by operating temperature, disk utilization, disk speed transition frequency, three energy-saving-related reliability affecting factors, PRESS estimates the reliability of entire disk array. Further, a new energy saving strategy with reliability awareness named Reliability and Energy Aware Distribution (READ) is developed in the light of the insights provided by PRESS.

Performance, Energy, and Reliability Balanced Dynamic Data Redistribution for Next Generation Disk Arrays

Contemporary disk arrays consist purely of hard disk drives, which normally provide huge storage capacities with low-cost and high-throughput for data-intensive applications. Nevertheless, they have some inherent disadvantages such as long access latencies, high annual disk replacement rates, fragile physical characteristics, and energy-inefficiency due to their build-in mechanical and electronic mechanisms. Flash-memory based solid state disks, on the other hand, although currently more expensive and inadequate in write cycles, offer much faster read accesses and are much more robust and energy efficient. To combine the complementary merits of hard disks and flash disks, in this research we developed a hybrid disk array architecture named HIT (hybrid disk storage) for data-intensive applications. Next, a dynamic data redistribution strategy called PEARL (performance, energy, and reliability balanced), which can periodically redistribute data between flash and hard disks to adapt to the changing data access patterns, is developed on top of the HIT architecture.

ECOS: An Energy-Efficient Cluster Storage System

Cluster storage systems are essential building blocks for many high-end computing infrastructures. Although energy conservation techniques have been intensively studied in the context of clusters and disk arrays, improving energy efficiency of cluster storage systems remains an open issue. To address this problem, we describe in this study an approach to implementing an energy-efficient cluster storage system or ECOS for short. ECOS relies on the architecture of cluster storage systems in which each I/O node manages multiple disks - one buffer disk and several data disks. Given an I/O node, the key idea behind ECOS is to redirect disk requests from data disks to the buffer disk. To balance I/O load among I/O nodes, ECOS might redirect requests from one I/O node into the others. Redirecting requests is a driving force of energy saving and the reason is two-fold. First, ECOS makes an effort to keep buffer disks active while placing data disks into standby in a long time period to conserve energy. Second, ECOS reduces the number of disk spin downs/ups in I/O nodes. The idea of ECOS was implemented in a Linux cluster, where each I/O node contains one buffer disk and two data disks. Experimental results show that ECOS improves the energy efficiency of traditional cluster storage systems where buffer disks are not employed. Adding one extra buffer disk into each I/O node seemingly has negative impact on energy saving. Interestingly, our results indicate that ECOS equipped with extra buffer disks is more energy efficient than the same cluster storage system without the buffer disks. The implication of the experiments is that using existing data disks in I/O nodes to perform as buffer disks can achieve even higher energy efficiency.

Performance Evaluation of Energy-Efficient Parallel I/O Systems with Write Buffer Disks

To conserve energy consumption in parallel I/O systems, one can immediately spin down disks when disk are idle; however, spinning down disks might not be able to produce energy savings due to penalties of spinning operations. Unlike powering up CPUs, spinning down and up disks need physical movements. Therefore, energy savings provided by spinning down operations must offset energy penalties of the disk spinning operations. To substantially reduce the penalties incurred by disk spinning operations, we developed a novel approach to conserving energy of parallel I/O systems with write buffer disks, which are used to accumulate small writes using a log file system. Data sets buffered in the log file system can be transferred to target data disks in a batch way. Thus, buffer disks aim to serve a majority of incoming write requests, attempting to reduce the large number of disk spinning operations by keeping data disks in standby for long period times. Interestingly, the write buffer disks not only can achieve high energy efficiency in parallel I/O systems, but also can shorten response times of write requests. To evaluate the performance and energy efficiency of our parallel I/O systems with buffer disks, we implemented a prototype using a cluster storage system as a testbed. Experimental results show that under light and moderate I/O load, buffer disks can be employed to significantly reduce energy dissipation in parallel I/O systems without adverse impacts on I/O performance.

HYBUD: An Energy-Efficient Architecture for Hybrid Parallel Disk Systems

Although flash memory is very energy-efficient compared to disk drives, flash memory is too expensive to use as a major component in large-scale storage systems. In other words, it is not a cost-effective way to make use of large flash memory to build energy-efficient storage systems. To address this problem, in this study we proposed a hybrid disk architecture or HYBUD that integrates a non-volatile flash memory with buffer disks to build cost-effective and energy-efficient parallel disk systems. While the most popular data sets are cached in flash memory, the second most popular data sets can be stored and retrieved from buffer disks. HYBUD is energy efficient because flash memory coupled with buffer disks can serve a majority of incoming disk requests, thereby keeping a large number of other data disks in the low-power state for longer period times. Furthermore, HYBUD is cost-effective by the virtue of inexpensive buffer disks assisting flash memory to cache a huge amount of popular data. Experimental results demonstratively show that compared with two existing non-hybrid architectures, HYBUD provides significant energy savings for parallel disk systems in a very cost effective way.

Energy-Aware Prefetching for Parallel Disk Systems

In this study we design and evaluate an energy-aware prefetching strategy for parallel disk systems consisting of a small number of buffer disks and large number of data disks. Using buffer disks to temporarily handle requests for data disks, we can keep data disks in the low-power mode as long as possible. Our prefetching algorithm aims to group many small idle periods in data disks to form large idle periods, which in turn allow data disks to remain in the standby state to save energy. To achieve this goal, we utilize buffer disks to aggressively fetch popular data from regular data disks into buffer disks, thereby putting data disks into the standby state for longer time intervals. A centrepiece in the prefetcing mechanism is an energy-saving prediction model, based on which we implement the energy-saving calculation module that is invoked in the prefetching algorithm. We quantitatively compare our energy-aware prefetching mechanism against existing solutions, including the dynamic power management strategy. Experimental results confirm that the buffer-disk-based prefetching can significantly reduce energy consumption in parallel disk systems by up to 50 percent. In addition, we systematically investigate the energy efficiency impact that varying disk power parameters has on our prefetching algorithm.

DORA: A Dynamic File Assignment Strategy with Replication

Compared with numerous static file assignment algorithms proposed in the literature, very few investigations on the dynamic file allocation problem have been accomplished. Moreover, none of them has integrated file replication techniques into file assignment algorithms in a highly dynamic file system where files are created or deleted on the fly and their access patterns varied over time. We argue that file replication and file assignment can act in concert to boost the performance of parallel disk systems. In this study, we propose a new dynamic file assignment strategy called DORA (dynamic round robin with replication). The advantages of DORA can be attributed to its two main characteristics. First, it takes the dynamic nature of file access patterns into account to adapt to a changing workload condition. Second, it utilizes file replication techniques to complement file assignment schemes so that system performance can be further improved. Experimental results demonstrate that DORA performs consistently better than existing algorithms.

Collaboration-Oriented Data Recovery for Mobile Disk Arrays

Mobile disk arrays, disk arrays located in mobile data centers, are crucial for mobile applications such as disaster recovery. Due to their unusual application domains, mobile disk arrays face several new challenges including harsh operating environments, very limited power supply, and extremely small number of spare disks. Consequently, data reconstruction schemes for mobile disk arrays must be performance-driven, reliability-aware, and energy-efficient. In this study, we develop a flash assisted data reconstruction strategy called CORE (collaboration-oriented reconstruction) on top of a hybrid disk array architecture, where hard disks and flash disks collaborate to shorten data reconstruction time, alleviate performance degradation during disk recovery. Experimental results demonstrate that CORE noticeably improves the performance and energy-efficiency over existing schemes.

A File Assignment Strategy Independent of Workload Characteristic Assumptions.

The problem of statically assigning non-partitioned files in a parallel I/O system has been extensively investigated. A basic workload characteristic assumption of most existing solutions to the problem is that there exists a strong inverse correlation between file access frequency and file size. In other words, the most popular files are typically small in size, while the large files are relatively unpopular. Recent studies on the characteristics of web proxy traces suggested, however, the correlation, if any, is so weak that it can be ignored. Hence, the following two questions arise naturally. First, can existing algorithms still perform well when the workload assumption does not hold? Second, if not, can one develop a new file assignment strategy that is immune to the workload assumption? To answer these questions, we first evaluate the performance of three well-known file assignment algorithms with and without the workload assumption, respectively. Next, we develop a novel static non-partitioned file assignment strategy for parallel I/O systems, called static round-robin (SOR), which is immune to the workload assumption. Comprehensive experimental results show that SOR consistently improves the performance in terms of mean response time over the existing schemes.

New course development

Our education activities include the development of a new course – Advanced Computer Security - for graduate students. This course was intended to give students a strong background in understanding design and development of secure systems in general and secure storage systems in particular.