Storage Latency Prediction via Concurrency Effect Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computing devices face limitations in providing efficient computer implemented services due to the constraints of their hardware and software components, making it impractical to predict and manage storage system performance under varying load conditions and configurations.
Innovation Solution
An information handling system that uses a concurrency effect model to estimate latency by analyzing access patterns and media access times, allowing for the prediction of storage system performance and optimization of resource management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If storage system performance is predicted under varying load conditions, then service quality and availability improve, but system complexity increases
Solution Approach 1:
The patent creates a virtual copy of the storage system's access patterns through tracing mechanisms. Instead of physically testing multiple storage configurations, the system copies real access patterns and applies them to virtual models for prediction, eliminating the need for complex physical experimentation while maintaining prediction accuracy.
Solution Approach 2:
The patent introduces an intermediary prediction layer between the storage system and the services it supports. This intermediary uses concurrency effect models and access pattern tracing to mediate performance predictions, avoiding direct complex interactions between varying load conditions and multiple storage configurations.
2Measurement precision
If concurrency effects are modeled accurately, then latency prediction precision improves, but computational requirements increase
Solution Approach 1:
The patent performs preliminary actions by tracing and recording access patterns during normal operation before prediction is needed. Concurrency effect models are pre-computed based on historical data, so when prediction is required, the system only needs to apply pre-existing models to new access patterns rather than performing complex computations from scratch.
Solution Approach 2:
The patent applies partial action by focusing computational efforts only on the specific concurrency effects that most impact latency prediction. Rather than modeling all possible system interactions, the system identifies and models only the critical concurrency patterns that significantly affect performance, reducing unnecessary computational overhead.
3Productivity
If storage systems are sized and reconfigured to suit workloads, then resource efficiency improves, but system adaptability requirements increase
Solution Approach 1:
The patent implements feedback by continuously monitoring actual storage performance and comparing it against predicted latency values. This feedback loop enables the system to learn from discrepancies between predicted and actual performance, automatically refining access pattern traces and concurrency models to improve future predictions and resource allocation decisions.
Solution Approach 2:
The patent enables parameter changes by allowing dynamic adjustment of storage system configuration based on workload characteristics. The system modifies parameters such as cache sizes, buffer allocations, and access patterns to optimize performance for different workload types, with the prediction system adapting to these changes through continuous learning.
Data Source
AI summary
An information handling system for managing a storage system includes storage and a storage manager. The storage is for storing a concurrency effect model for the storage system. The storage manager obtains a request for a latency prediction for the storage system; obtains media access time estimates for the storage system using a trace that specifies: a series of accesses; and responses, by the storage system, to the series of accesses; obtain concurrency estimates for the trace using the media access time estimates; obtains modifications for the media access time estimates using: the concurrency effect model; and the concurrency estimates; updates the media access time estimates using the modifications to obtain a latency prediction for the storage system; and performs an action set, based on the latency prediction, to manage operation of the storage system.