Xiao Qin's Research
Final Report
BUD: A Buffer-Disk Architecture for Energy Conservation in Parallel Disk Systems
Findings
An Energy-Efficient Framework for Large-Scale Parallel Storage Systems
A Simulation Framework for Energy-efficient Data Grids
An Energy-Efficient Scheduling Algorithm Using Dynamic Voltage Scaling
for Parallel Applications on Clusters
Load-Balancing Strategies for Energy-Efficient Parallel Storage Systems with Buffer Disks
Sacrificing Reliability for Energy Saving: Is It Worthwhile for Disk Arrays
Load-Balancing Strategies for Energy-Efficient Parallel Storage Systems with Buffer Disks
Energy-Efficient Prefetching for Parallel I/O Systems with Buffer Disks
Improving Reliability and Energy Efficiency of Disk Systems via Utilization Control
Energy Conservation for Real-Time Disk Systems with I/O Burstiness
An Adaptive Energy-Conserving Strategy for Parallel Disk Systems
DARAW: A New Write Buffer to Improve Parallel I/O Energy-Efficiency
MICRO: A Multi-level Caching-based Reconstruction Optimization for Mobile Storage Systems
PEARL: Performance, Energy, and Reliability Balanced Dynamic Data Redistribution for Next Generation Disk Arrays
HyBUD: An Energy-Efficient Architecture for Hybrid Parallel Disk Systems
Energy-Aware Prefetching for Parallel Disk Systems
DORA: A Dynamic File Assignment Strategy with Replication
Collaboration-Oriented Data Recovery for Mobile Disk Arrays
A File Assignment Strategy Independent of Workload Characteristic Assumptions
References
[Ruan et al., 2009] 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 work we propose a novel strategy, which aims to significantly conserve energy while reducing average I/O response times. This goal is 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.
The Figure below shows the energy consumption and average response time of a parallel disk system with DARAW and the same disk system without DARAW. The results indicate that when we increase SRB, more energy can be saved. The results were expected since when the SRB grows, the system can write more requests into data disks with reduced number of power state transitions. However, we also observe that when the SRB equals to one, the energy consumption is even greater than the disk system without DARAW. This interesting tend can be explained as follows. Our parallel disk system has a buffer-disk layer that also consumes energy. If there is insufficient number of requests written into a data disk when a power-state transition occurs, energy conserved cannot offset energy overhead introduced by the buffer disk. When we did the experiment with a trace generated by increasing values of λ, we observe that energy consumptions in both the non-DARAW parallel disk system and the system with DARAW decrease.
IBM 40GNX Travelstar. Energy Consumption and Average Response Time Compare
Note that all the traces have the same number of disk requests. This implies the fact that when λ is high, all requests are arriving at the system within a shorter period of time, making all the disks stay in the active state for a shortened time interval. This is the reason behind the result that energy consumption of the system with DARAW when λ is set to 0.02 is slightly smaller than that of the system when λ is 0.01. However, the power consumption of the non-DARAW disk system is significantly smaller when λ is 0.01 as compared to λ = 0.02. Once the arrival rate goes up, each data disk in the non-DARAW system has greater probability to receive a request when it is working. Thus, the number of power-state transitions can be noticeably reduced. When λ is set to 0.02, there is less of an opportunity to simultaneously save energy and satisfy response times. When we increase the number of buffer disks from 5 to 20, DARAW can conserve energy while guaranteeing reasonably short response times. An appealing result shown in the above Figure is that compared with the parallel I/O system without DARAW, our approach not only achieves significant energy savings, but also reduces response times. In DARAW, the response time is the time when a request is written in to a data or buffer disk. Since buffer disks can serve coming requests when data disks are sleeping, the response time can be noticeably shortened.
Our results show that DARAW works well for parallel I/O systems with both high performance disks and mobile disks. DARAW achieves promising results when the arrival rate is low. When the request arrival rate rises, we can either use high-performance hard drives or add more buffer disks to boost I/O performance. If the arrival rate is high, all data disks are busy serving requests, leaving no opportunity to save energy. As the SRB parameter grows, DARAW is given a greater window of opportunity to conserve energy. However, if the SRB is too large, it may cause a “traffic jam” inside the parallel I/O system with buffer disks.
[Xie and Sun, 2008a] Mainstream energy conservation schemes for disk arrays inherently affect the reliability of disks. A thorough understanding of the relationship between energy saving techniques and disk reliability is still an open problem, which prevents effective design of new energy saving techniques and application of existing approaches in reliability-critical environments. As one step towards solving this problem, we investigated an empirical reliability model, called Predictor of Reliability for Energy Saving Schemes (PRESS). The architecture of the PRESS model is given below:
Fed by three energy-saving-related reliability-affecting factors, operating temperature, utilization, and disk speed transition frequency, PRESS estimates the reliability of entire disk array. In what follows, we present two 3-dimennsional figures to represent the PRESS model at operating temperature 40 C (Figure 5a) and 50 C (Figure 5b), respectively.
Further, we developed a new energy saving strategy with reliability awareness called Reliability and Energy Aware Distribution (READ) is developed in the light of the insights provided by PRESS. Experimental results demonstrate that compared with existing energy saving schemes, MAID and PDC, READ consistently performs better in performance and reliability while achieving a comparable level of energy consumption.
