Configurable Data Mover for Multi-Tier Storage Power Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional storage tiering in high-performance computer systems leads to excessive power consumption, limiting computational performance due to inefficient data movement between storage tiers.
Innovation Solution
Implementing a multi-tier storage system with dynamic power management using configurable data mover modules that alter data movement based on a power management configuration file, optimizing data transfer modes and schedules to minimize power consumption during compute phases and accelerate IO operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data is continuously moved between storage tiers to optimize performance, then IO performance is improved, but power consumption increases excessively
Solution Approach 1:
The patent implements periodic data movement operations synchronized with compute phases. Data is moved between storage tiers in periodic bursts during idle periods rather than continuously, reducing power consumption while maintaining necessary data availability for computational workloads.
Solution Approach 2:
The data movement system dynamically adjusts its operation based on computational phase state. During active compute phases, data movement is minimized or paused to reduce power consumption. During idle phases or transitions, data movement operations are performed to prepare data for upcoming compute phases, creating a dynamic adaptation to system state.
2Reliability
If data movement operations are performed during compute phases to maintain data availability, then IO performance is maintained, but power consumption increases and limits computational performance
Solution Approach 1:
Data is pre-positioned in optimal storage tiers before compute phases begin. The system performs data movement operations during idle periods to ensure required data is available in fast storage before computational workloads start, eliminating the need for additional data movement during power-sensitive compute phases.
Solution Approach 2:
The system monitors computational phase state and uses this feedback to control data movement operations. When compute phases are detected, data movement is automatically reduced or paused. When idle phases occur, data movement is activated to maintain data availability, creating a closed-loop control system that responds to real-time conditions.
3Speed
If multiple parallel threads are used for data movement to accelerate IO operations, then data transfer speed is improved, but power consumption increases
Solution Approach 1:
The system uses a single data movement thread during idle phases rather than multiple parallel threads, providing sufficient data transfer capability while consuming less power. Multiple parallel threads are reserved for use during critical periods when maximum throughput is required, accepting that these high-power operations occur only occasionally during idle time.
Data Source
AI summary
An apparatus in one embodiment comprises a storage system having at least first and second storage tiers each comprising a plurality of storage devices. The storage system implements at least one configurable data mover module adapted to interact with the first and second storage tiers and comprising at least one power management configuration file. The data mover module alters a manner in which data is moved between the first and second storage tiers based at least in part on a change in state of the power management configuration file. For example, the data mover module may be configurable via the power management configuration file in a mode of operation in which data movement from the first storage tier to the second storage tier occurs for data files utilized by a given application only after completion of a compute phase for that application.