Adaptive Storage Routing via Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High latency in data operations within multi-tiered storage systems hampers the efficiency of computer system operations, as existing systems lack an adaptive method to optimize routing of service requests across different storage tiers based on their states and future expected benefits.
Innovation Solution
A prediction architecture using Markov Decision Process and reinforcement learning algorithms to predict the benefits of each storage tier, routing service requests to maximize the total expected sum of benefits, thereby reducing latency by adaptively tuning the routing algorithm based on the state transitions over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a fixed set of criteria is used to route incoming requests in multi-tiered storage systems, then the routing process is simple and deterministic, but the system cannot adapt to changing storage tier states and future expected benefits, resulting in suboptimal latency reduction
Solution Approach 1:
The patent implements dynamic routing by transitioning from fixed criteria to adaptive algorithms that continuously adjust based on real-time storage tier states and predicted future benefits. The system dynamically selects routing strategies (e.g., shortest path, most beneficial path) based on current conditions, enabling the routing mechanism to adapt to changing states while managing complexity through structured decision frameworks.
Solution Approach 2:
The patent employs feedback mechanisms where the system monitors actual performance metrics (latency, throughput) and uses this information to refine routing decisions. The prediction architecture receives feedback about storage tier states and adjusts future routing predictions accordingly, creating a closed-loop system that continuously improves adaptability based on observed behavior and outcomes.
2Reliability
If storage systems use higher capacity disk drives and multi-tiered architectures, then data integrity and capacity increase, but latency in data operations increases
Solution Approach 1:
The patent segments the storage system into multiple tiers with different characteristics (capacity, speed, reliability) and routes requests to specific segments based on predicted benefits. By dividing the storage architecture into distinct tiers and applying selective routing, the system can access data through optimal paths that balance integrity requirements with latency constraints, rather than treating storage as a monolithic resource.
Solution Approach 2:
The prediction architecture performs preliminary analysis of storage tier states and future expected benefits before routing decisions are made. By predicting future states and pre-determining optimal routing paths, the system can prepare routing decisions in advance, reducing the time required for actual data operations and minimizing latency while maintaining data integrity through proactive rather than reactive routing.
3Loss of time
If the routing algorithm adaptively tunes functions based on state transitions, then latency reduction is achieved, but the computational overhead and complexity of the prediction architecture increases
Solution Approach 1:
The patent changes routing parameters dynamically based on predicted storage tier states and future benefits. Instead of using fixed routing parameters, the system adjusts parameters such as preferred paths, tier selections, and routing priorities based on real-time conditions. This parameter adaptation enables latency reduction by optimizing routing decisions for each specific state transition scenario while managing complexity through parameterized rather than structurally complex approaches.
Data Source
AI summary
A method for routing an incoming service request is described wherein the service request is routed to a selected storage tier based on that selected storage tier having a predicted value indicating a state having greater utility as compared with the predicted value of the state associated with at least one other storage tier within the storage system. A computer system comprising a multi-tier storage system is described, the multi-tier storage system having a routing algorithm configured to adaptively tune functions which map variables describing the state of each storage tier of the storage system into the average latency experienced by incoming service requests associated with the storage tier.