How reliable are parallel disk systems when energy-saving schemes are involved?

SUMMARYThe concept of green storage in cluster computing has recently attracted enormous interest among researchers. Consequently, several energy-efficient solutions, such as multi-speed disks and disk spin down methods, have been proposed to conserve power in storage systems and improve disk access. Some researchers have assessed their proposed solutions via simulations, while others have used real-world experiments. Both methods have advantages and disadvantages. Simulations can more swiftly assess the benefits of energy-efficient solutions, but various measurement errors can arise from procedural shortcomings. For instance, many power simulation tools fail to consider how heat increases the power overhead of disk operations. Some researchers claim that their modeling methods reduce the measurement error to 5% in the single disk model. However, the demand for large-scale storage systems is growing rapidly. Traditional power measurement using a single disk model is unsuited to such systems because of their complex storage architecture and the unpredictability of numerous disks. Consequently, a number of studies have conducted real machine experiments to assess the performance of their solutions in terms of power conservation, but such experiments are time consuming. To address this problem, this study proposes an efficient simulation tool called Benchmark Analysis Software for Energy-efficient Solution (BASE), which can accurately estimate disks' power consumption in large-scale storage systems. We evaluate the performance of BASE on real-world traces of Academia Sinica (Taiwan) and Florida International University. BASE incorporates an analytical method for assessing the reliability of energy-efficient solutions. The analytical results demonstrate that the measurement error of BASE is 2.5% lower than that achieved in real-world experiments involving energyestimation experiments. Moreover, the results of simulations to assess solution reliability are identical to those obtained through real-world experiments.

show abstract

“…

A (f) = 1.51 {normale}^{- 6} f^{2} - 1.09 {normale}^{- 5} f + 1.39 {normale}^{- 2},

where f denotes the power‐state transition frequency. The reliability model was proposed by Yin et al .…”

Section: The Design Of Benchmark Analysis Software For Energy‐efficiementioning

confidence: 99%

BASE: an assistant tool to precisely simulate energy consumption and reliability of energy‐efficient storage systems

2015

View full text Add to dashboard Cite

show abstract

“…The service type indicates whether the archive system can be accessed by all users or only by a specific user/group of users. And in most cases, researchers did not verify the reliability of their solutions, with only CacheRAID and Yin et al's approach [11] using the MINT model to analyze the reliability of the proposed energy-efficient solutions.…”

Section: A Energy-efficient Schemesmentioning

confidence: 99%

“…Most of these solutions use the disk spin down mechanism as the fundamental technology for energy-saving in a system. However, the disk spin-down mechanism also increases risk of data loss, because the disk frequently changes its power-state [10], [11]. Data loss is the worst situation for all providers of data warehouse services, so most service providers use at least one fault tolerance mechanism for data in their data center.…”

Section: Introductionmentioning

confidence: 99%

A study of suitable fault tolerance frameworks for an energy-efficient storage system

Chen

Wei

et al. 2015

2015 International Conference on Collaboration Technologies and Systems (CTS)

View full text Add to dashboard Cite

A new kind of cloud service, called Database-asaservice (DaaS), provides a novel paradigm which allows data to be stored in remote databases. The types of database used in remote sites can be divided into two categories: Relational Databases (RDB) and NoSQL. NoSQL databases store data in a flexible format, and without schema, so NoSQL databases can be easily scaled across multiple devices. Because of the advantages of the NoSQL databases, many large data applications use NoSQL databases as the backend data management system. Because NoSQL databases usually access the stored data as files, through a distributed file system, the problem of decentralized data management is critical in NoSQL databases. In order to address this issue, many energy-efficient algorithms for file placement are used in distributed file systems. However, these energy-efficient algorithms only focus on reducing the power consumption of storage systems and they ignore the cost of data reliability, which is affected by frequently setting the disk status to standby mode. To maintain data reliability, energy-efficient storage systems usually combine framework of power conservation and a fault tolerance mechanism. However, most developers of energyefficient frameworks do not consider whether the fault tolerance mechanism can avoid data loss when this combination is used for energy-efficient storage systems. This study proposes a general energy-efficient framework that can be integrated different fault tolerance mechanisms. The experimental results show that the capability of the proposed framework with different erasure code algorithms.

show abstract

“…MINT is a framework developed to model reliability of parallel disk systems employing energy conservation techniques [19]. In the MINT framework, we studied the reliability impacts of two wellknown energy-saving techniques -the Popular Disk Concentration technique (PDC) and the Massive Array of Idle Disks (MAID).…”

Section: A Overviewmentioning

confidence: 99%

“…1) We studied two reliability model for Popular Data Concentration scheme (PDC) and Massive Array of Idle Disks (MAID) based on Mathematical Reliability Models for Energy-efficient Parallel Disk System (MINT) [19]; 2) We built a disk swapping mechanism to improve reliability of various load skewing techniques. 3) We studied the impacts of the disk swapping schemes on the reliability of PDC and MAID.…”

Section: Introductionmentioning

confidence: 99%

Improving reliability of energy-efficient parallel storage systems by disk swapping

Yin

Ruan

Manzanares

et al. 2009

2009 IEEE 28th International Performance Computing and Communications Conference

Self Cite

View full text Add to dashboard Cite

Abstract-The Popular Disk Concentration (PDC) technique and theMassive Array of Idle Disks (MAID) technique are two effective energy saving schemes for parallel disk systems. The goal of PDC and MAID is to skew I/O load towards a few disks so that other disks can be transitioned to low power states to conserve energy. I/O load skewing techniques like PDC and MAID inherently affect reliability of parallel disks because disks storing popular data tend to have high failure rates than disks storing cold data. To achieve good tradeoffs between energy efficiency and disk reliability, we first present a reliability model to quantitatively study the reliability of energy-efficient parallel disk systems equipped with the PDC and MAID schemes. Then, we propose a novel strategydisk swapping-to improve disk reliability by alternating disks storing hot data with disks holding cold data. We demonstrate that our diskswapping strategies not only can increase the lifetime of cache disks in MAID-based parallel disk systems, but also can improve reliability of PDC-based parallel disk systems.

show abstract

How reliable are parallel disk systems when energy-saving schemes are involved?

Cited by 13 publications

References 21 publications

BASE: an assistant tool to precisely simulate energy consumption and reliability of energy‐efficient storage systems

BASE: an assistant tool to precisely simulate energy consumption and reliability of energy‐efficient storage systems

A study of suitable fault tolerance frameworks for an energy-efficient storage system

Improving reliability of energy-efficient parallel storage systems by disk swapping

Contact Info

Product

Resources

About