Dynamic Load Balancing for Hybrid SCM and NAND Storage Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In storage systems utilizing both high-speed SCM SSDs and low-speed NAND SSDs, existing load balancing methods fail to effectively utilize the high-speed performance of SCM SSDs due to increased latency when the SCM SSD is under intensive load, leading to decreased performance.
Innovation Solution
A method where a master server collects load information from multiple servers, determines the load state of the SCM SSD, and dynamically transfers data between SCM SSD and NAND SSD based on the load state, ensuring that data is written or read from the faster device when possible, thereby distributing the load and maintaining high access speeds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If load balancing control is performed in a storage system with SCM SSD and NAND SSD, then data is transferred between storage devices according to access frequency, but the latency increases and high-speed performance of SCM cannot be utilized effectively
Solution Approach 1:
The patent implements dynamic load control by continuously monitoring the load state of the SCM SSD and adjusting data placement decisions in real-time. When the SCM load exceeds a threshold, the system dynamically changes behavior to write data directly to the NAND SSD, preventing latency degradation. This dynamic adaptation resolves the contradiction by maintaining high-speed access characteristics only when necessary.
Solution Approach 2:
The system changes the operational parameters of the storage system based on the SCM's load state. By monitoring metrics such as queue depth and access patterns, the system transitions between different operational modes: utilizing the SCM for high-speed access when lightly loaded, and bypassing it when heavily loaded. This parameter-based control ensures optimal performance by adapting to changing conditions.
2Speed
If data is always written to and read from the high-speed SCM SSD, then access speed is improved, but when the SCM SSD is under intensive load, the latency increases and performance decreases
Solution Approach 1:
The patent implements a feedback mechanism that continuously monitors the load state of the SCM SSD and uses this information to control data placement decisions. The load determination unit receives feedback about queue depth, access patterns, and current utilization, then adjusts the data writing control accordingly. This feedback loop maintains performance consistency by preventing the SCM from operating in high-latency states.
Solution Approach 2:
The storage system performs self-regulation by automatically monitoring its own load state and making intelligent decisions about data placement without external intervention. The controller assesses its own performance characteristics and adjusts operations to maintain optimal speed while avoiding latency degradation, enabling the system to serve itself and maintain reliable performance.
3Quantity of substance
If hierarchical control transfers data from SCM to NAND based on access frequency, then storage capacity is optimized, but the complexity of control increases and high-speed performance is compromised
Solution Approach 1:
The patent extracts the load monitoring and decision-making function from complex hierarchical control and implements a simplified approach focused specifically on load state assessment. Rather than implementing full hierarchical control with multiple levels of decision-making, the system extracts the essential function of monitoring SCM load and making binary decisions about data placement, significantly reducing control complexity while maintaining effective capacity utilization.
Data Source
AI summary
An information processing apparatus includes a memory and a processor and accesses a first storage device and a second storage device wherein an access speed of the second storage device is higher than an access speed of the first storage device. The memory stores information relating to a request in a request from the information processing apparatus to the second storage device. The processor, which is connected to the memory, determines a load on the second storage device based on the information relating to the request.