Muutke küpsiste eelistusi

Dynamic Resource Allocation in Embedded, High-Performance and Cloud Computing [Kõva köide]

Edited by , Edited by , Edited by
Teised raamatud teemal:
Dynamic Resource Allocation in Embedded, High-Performance and Cloud ComputingSuurem pilt
Teised raamatud teemal:
Püsilink: https://www.kriso.ee/db/9788793519084.html
The availability of many-core computing platforms enables a wide variety of technical solutions for systems across the embedded, high-performance and cloud computing domains. However, large scale manycore systems are notoriously hard to optimise. Choices regarding resource allocation alone can account for wide variability in timeliness and energy dissipation (up to several orders of magnitude).

Dynamic Resource Allocation in Embedded, High-Performance and Cloud Computing covers dynamic resource allocation heuristics for manycore systems, aiming to provide appropriate guarantees on performance and energy efficiency. It addresses different types of systems, aiming to harmonize the approaches to dynamic allocation across the complete spectrum between systems with little flexibility and strict real-time guarantees all the way to highly dynamic systems with soft performance requirements.

Technical topics presented in the book include:
• Load and Resource Models
• Admission Control
• Feedback-based Allocation and Optimisation
• Search-based Allocation Heuristics
• Distributed Allocation based on Swarm Intelligence
• Value-Based Allocation

Each of the topics is illustrated with examples based on realistic computational platforms such as Network-on-Chip manycore processors, grids and private cloud environments.
Preface xi
Acknowledgements xiii
List of Figures
xv
List of Tables
xix
List of Algorithms
xxi
List of Abbreviations
xxiii
1 Introduction
1(10)
1.1 Application Domains
2(2)
1.2 Related Work
4(3)
1.2.1 Allocation Techniques for Guaranteed Performance
4(2)
1.2.2 Allocation Techniques for Energy-efficiency
6(1)
1.3 Challenges
7(4)
1.3.1 Load Representation
7(1)
1.3.2 Monitoring and Feedback
8(1)
1.3.3 Allocation of Modal Applications
8(1)
1.3.4 Distributed Allocation
9(1)
1.3.5 Value-based Allocation
9(2)
2 Load and Resource Models
11(14)
2.1 Related Work
12(1)
2.2 Requirements
13(4)
2.2.1 Requirements on Modelling Load Structure
13(1)
2.2.1.1 Singleton
14(1)
2.2.1.2 Independent jobs
14(1)
2.2.1.3 Single-dependency jobs
14(1)
2.2.1.4 Communicating jobs
14(1)
2.2.1.5 Multi-dependency jobs
14(1)
2.2.2 Requirements on Modelling Load Temporal Behaviour
14(1)
2.2.2.1 Single appearance
15(1)
2.2.2.2 Strictly periodic
15(1)
2.2.2.3 Sporadic
15(1)
2.2.2.4 Aperiodic
15(1)
2.2.2.5 Fully dependent
15(1)
2.2.2.6 N out of M dependent
15(1)
2.2.3 Requirements on Modelling Load Resourcing Constraints
15(1)
2.2.3.1 Untyped job
16(1)
2.2.3.2 Single-typed job
16(1)
2.2.3.3 Multi-typed job
16(1)
2.2.4 Requirements on Modelling Load Characterisation
16(1)
2.2.4.1 Fixed load
16(1)
2.2.4.2 Probabilistic load
16(1)
2.2.4.3 Typed fixed load
16(1)
2.2.4.4 Typed probabilistic load
16(1)
2.3 An Interval Algebra for Load and Resource Modelling
17(6)
2.3.1 Modelling Load Structure
19(1)
2.3.2 Modelling Load Temporal Behaviour
19(1)
2.3.3 Modelling Load Resourcing Constraints
20(2)
2.3.4 Modelling Load Characterisation
22(1)
2.3.5 Stochastic Time
22(1)
2.4 Summary
23(2)
3 Feedback-Based Admission Control Heuristics
25(26)
3.1 System Model and Problem Formulation
26(1)
3.1.1 Platform Model
26(1)
3.1.2 Application Model
27(1)
3.2 Distributed Feedback Control Real-Time Allocation
27(2)
3.3 Experimental Results
29(6)
3.3.1 Controller Tuning
29(3)
3.3.2 Stress Tests
32(2)
3.3.3 Random Workloads
34(1)
3.4 Dynamic Voltage Frequency Scaling
35(2)
3.5 Applying Controllers to Steer DVFS
37(3)
3.6 Experimental Results
40(6)
3.6.1 Controller Tuning
40(2)
3.6.2 Random Workloads
42(4)
3.7 Related Work
46(2)
3.8 Summary
48(3)
4 Feedback-Based Allocation and Optimisation Heuristics
51(22)
4.1 System Model and Problem Formulation
52(2)
4.1.1 Application Model
53(1)
4.1.2 Platform Model
53(1)
4.1.3 Problem Formulation
54(1)
4.2 Performing Runtime Admission Control and Load Balancing to Cope with Dynamic Workloads
54(4)
4.3 Experimental Results
58(12)
4.3.1 Number of Executed Tasks, Rejected Tasks and Schedulability Tests
60(1)
4.3.1.1 Periodic workload
60(2)
4.3.1.2 Random workload
62(2)
4.3.2 Dynamic Slack, Setpoint and Controller Output
64(1)
4.3.2.1 Periodic workload
64(2)
4.3.2.2 Light workload
66(1)
4.3.3 Core Utilization
66(2)
4.3.4 Case Study: Industrial Workload Having Dependent Jobs
68(2)
4.4 Related Work
70(2)
4.5 Summary
72(1)
5 Search-Based Heuristics for Modal Application
73(22)
5.1 System Model and Problem Formulation
74(4)
5.1.1 Application Model
74(1)
5.1.2 Platform Model
75(2)
5.1.3 Problem Formulation
77(1)
5.2 Proposed Approach
78(13)
5.2.1 Mode Detection/Clustering
78(1)
5.2.2 Spanning Tree Construction
79(1)
5.2.3 Static Mapping for Initial Mode
80(2)
5.2.4 Static Mapping for Non-Initial Modes
82(2)
5.2.5 Schedulability Analysis for Taskset During Mode Changes
84(6)
5.2.6 On-Line Steps
90(1)
5.3 Related Works
91(2)
5.4 Summary
93(2)
6 Swarm Intelligence Algorithms for Dynamic Task Reallocation
95(24)
6.1 System Model and Problem Formulation
96(3)
6.1.1 Load Model
96(2)
6.1.2 Platform Model
98(1)
6.1.3 Problem Statement
98(1)
6.2 Swarm Intelligence for Resource Management
99(9)
6.2.1 PS -- Pheromone Signalling Algorithm
99(3)
6.2.2 PSRM -- Pheromone Signalling Supporting Load Remapping
102(6)
6.3 Evaluation
108(8)
6.3.1 Experiment Design
108(1)
6.3.1.1 Metrics
109(1)
6.3.1.2 Baseline Remapping Techniques
110(1)
6.3.2 Experimental Results
110(1)
6.3.2.1 Comparison between clustered approaches
110(1)
6.3.2.2 Comparison regarding video processing performance
111(1)
6.3.2.3 Comparison regarding communication overhead
112(1)
6.3.2.4 Comparison regarding processor utilisation
113(2)
6.3.3 Outlook
115(1)
6.4 Summary
116(3)
7 Value-Based Allocation
119(16)
7.1 System Model and Problem Formulation
120(3)
7.1.1 Many-Core HPC Platform Model
120(1)
7.1.2 Job Model
121(1)
7.1.3 Value Curve of a Job
121(1)
7.1.4 Energy Consumption of a Job
122(1)
7.1.5 Problem Formulation
122(1)
7.2 The Solution
123(5)
7.2.1 Profiling Based Approach (PBA)
123(2)
7.2.2 Non-profiling Based Approach (NBA)
125(3)
7.3 Evaluations
128(5)
7.3.1 Experimental Baselines
129(1)
7.3.2 Value and Energy Consumption at Different Arrival Rates
130(1)
7.3.3 Value and Energy Consumption with Varying Number of Nodes
131(1)
7.3.4 Value and Energy Consumption with Varying Number of Cores in Each Node
131(1)
7.3.5 Percentage of Rejected Jobs
132(1)
7.4 Related Works
133(1)
7.5 Summary
134(1)
References 135(16)
About the Authors 151
Leando Soares Indrusiak, Piotr Dziurzanski, Amit Kumar Singh