Building a columnar database on RAMCloud database design for the low-latency enabled data center /
This book examines the field of parallel database management systems and illustrates the great variety of solutions based on a shared-storage or a shared-nothing architecture.
| Main Author: | Tinnefeld, Christian, |
|---|---|
| Other Authors: | SpringerLink (Online service) |
| Format: | eBook |
| Language: | English |
| Published: |
Cham :
Springer,
[2015]
|
| Physical Description: |
1 online resource (xix, 130 pages) : illustrations. |
| Series: |
In-memory data management research.
|
| Subjects: |
| Summary: |
This book examines the field of parallel database management systems and illustrates the great variety of solutions based on a shared-storage or a shared-nothing architecture. Constantly dropping memory prices and the desire to operate with low-latency responses on large sets of data paved the way for main memory-based parallel database management systems. However, this area is currently dominated by the shared-nothing approach in order to preserve the in-memory performance advantage by processing data locally on each server. The main argument this book makes is that such an unilateral development will cease due to the combination of the following three trends: a) Today's network technology features remote direct memory access (RDMA) and narrows the performance gap between accessing main memory on a server and of a remote server to and even below a single order of magnitude. B) Modern storage systems scale gracefully, are elastic, and provide high-availability. C) A modern storage system such as Stanford's RAMCloud even keeps all data resident in the main memory. Exploiting these characteristics in the context of a main memory-based parallel database management system is desirable. The book demonstrates that the advent of RDMA-enabled network technology makes the creation of a parallel main memory DBMS based on a shared-storage approach feasible. |
|---|---|
| Item Description: |
Includes bibliographical references. Machine generated contents note: 1. Introduction -- 1.1. Motivation -- 1.2. Research Questions and Scope -- 1.3. Outline -- 2. Related Work and Background -- 2.1. Current Computing Hardware Trends -- 2.1.1. Larger and Cheaper Main Memory Capacities -- 2.1.2. Multi-Core Processors and the Memory Wall -- 2.1.3. Switch Fabric Network and Remote Direct Memory Access -- 2.2. In-Memory Database Management Systems -- 2.2.1. Column-and Row-Oriented Data Layout -- 2.2.2. Transactional Versus Analytical Versus Mixed Workload Processing -- 2.2.3. State-of-the-Art In-Memory Database Management Systems -- 2.3. Parallel Database Management Systems -- 2.3.1. Shared-Memory Versus Shared-Disk Versus Shared-Nothing -- 2.3.2. State-of-the-Art Parallel Database Management Systems -- 2.3.3. Database-Aware Storage Systems -- 2.3.4. Operator Placement for Distributed Query Processing -- 2.4. Cloud Storage Systems -- 2.4.1. State-of-the-Art Cloud Storage Systems -- 2.4.2. Combining Database Management and Cloud Storage Systems -- 2.5. Classification -- pt. I Database System Architecture for a Shared Main Memory-Based Storage -- 3. System Architecture -- 3.1. System Architecture -- Requirements, Assumptions, and Overview -- 3.2. AnalyticsDB -- 3.3. Stanford's RAMCloud -- 4. Data Storage -- 4.1. Mapping from Columnar Data to RAMCloud Objects -- 4.2. Main Memory Access Costs and Object Sizing -- 5. Data Processing -- 5.1. Database Operators in AnalyticsDB -- 5.2. Operator Push-Down into RAMCloud -- 5.3. From SQL Statement to Main Memory Access -- pt. II Database Operator Execution on a Shared Main Memory-Based Storage -- 6. Operator Execution on One Relation -- 6.1. Evaluating Operator Execution Strategies -- 6.2. Optimizing Operator Execution -- 6.3. Implications of Data Partitioning -- 7. Operator Execution on Two Relations -- 7.1. Grace Join -- 7.2. Distributed Block Nested Loop Join -- 7.3. Cyclo Join -- 7.4. Join Algorithm Comparison -- 7.5. Parallel Join Executions -- pt. III Evaluation -- 8. Performance Evaluation -- 8.1. Analytical Workload: Star Schema Benchmark -- 8.2. Mixed Workload: Point-of-Sales Customer Data -- 9. High-Availability Evaluation -- 10. Elasticity Evaluation -- pt. IV Conclusions -- 11. Conclusions. This book examines the field of parallel database management systems and illustrates the great variety of solutions based on a shared-storage or a shared-nothing architecture. Constantly dropping memory prices and the desire to operate with low-latency responses on large sets of data paved the way for main memory-based parallel database management systems. However, this area is currently dominated by the shared-nothing approach in order to preserve the in-memory performance advantage by processing data locally on each server. The main argument this book makes is that such an unilateral development will cease due to the combination of the following three trends: a) Today's network technology features remote direct memory access (RDMA) and narrows the performance gap between accessing main memory on a server and of a remote server to and even below a single order of magnitude. B) Modern storage systems scale gracefully, are elastic, and provide high-availability. C) A modern storage system such as Stanford's RAMCloud even keeps all data resident in the main memory. Exploiting these characteristics in the context of a main memory-based parallel database management system is desirable. The book demonstrates that the advent of RDMA-enabled network technology makes the creation of a parallel main memory DBMS based on a shared-storage approach feasible. |
| Physical Description: |
1 online resource (xix, 130 pages) : illustrations. |
| Bibliography: |
Includes bibliographical references. |
| ISBN: |
9783319207117 3319207113 |
| ISSN: |
2196-8055. |