Support SLA via Adaptive Mapping and HeterogeneousStorage Devices in Bigdata Distributed Storage Systems = 빅데이터 분산 스토리지 시스템에서 적응적 매핑과 이기종 스토리지를 활용한 SLA 보장 기법

다국어 초록 (Multilingual Abstract)

Distributed storage system also known as cloud storage solution was created to provide scalability and reliability to store large and complex amount of bigdata. In addition to the functionality of distributed storage system, QoS (Quality of Service) especially to support different type of clients ranging from urgent time-critical to best-effort is one of criteria to gain more popularity and become a vital component in a distributed storage system.

We proposes a new resource management scheme that supports SLA (Service Level
Agreement) in a bigdata distributed storage system. Basically, it makes use of two modes,isolated mode and shared mode, in an adaptive manner. In specific, to ensure different QoS requirements among clients, it isolates storage devices into two regions, one for urgent clients and the other for normal clients. When there is no urgent client, it switches to the shared mode so that normal clients can access all storage devices, thus achieving full performance. To provide this adaptive mapping effectively, we devise two techniques, called logical cluster and normal inclusion.

In addition, we explore how to exploit heterogeneous storage devices, HDDs and SSDs, for supporting SLA. We observe that separating data and metadata into different devices gives a positive impact on the performance per price ratio. Real implementation-based evaluation results show that our proposal can prevent urgent clients from being interfered by normal clients while outperforms a fixed mapping based scheme.

목차 (Table of Contents)

• I. Introduction
• II. Background
• 2.1 Ceph Architecture
• 2.2 Mapping Mechanism
• 2.3 Motivation
• III. Support SLA
• 3.1 Adaptive Mapping
• 3.2 Heterogeneous Storage Devices
• IV. Evaluation
• 4.1 Experiment Environment
• 4.2 Mapping Configuration
• 4.3 OSD Configuration
• 4.4 Effective of Adaptive Mapping
• 4.5 Effect of Heterogeneous Storage Devices
• V. Related Works
• VI. Conclusions
• References