Controller Bypassing Cache via Moving Average Throughput
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current storage device controllers, such as RAID controllers, face challenges in accurately and efficiently bypassing the cache to reduce latency and increase performance due to bandwidth limitations of the cache bus, as existing firmware is slow to detect when to bypass the cache and may not accurately determine the need for bypassing, leading to potential bottlenecks.
Innovation Solution
A method is introduced where a threshold is set based on the cache bus bandwidth and storage device type, and a moving average of throughput is measured between the storage device and host, allowing the controller to bypass the cache and send requests directly to the storage device when the moving average equals the threshold, thereby determining when to bypass the cache more quickly and accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the controller uses existing firmware to detect when to bypass the cache, then the system can attempt to reduce latency, but the detection is slow and inaccurate, causing potential bandwidth bottlenecks
Solution Approach 1:
The patent changes the detection parameter from complex firmware-based multi-factor analysis to a simple moving average of throughput measurements. By using moving average calculations on throughput data, the system achieves fast and accurate detection of cache bus bandwidth saturation, resolving the contradiction between detection accuracy and detection speed.
2Speed
If the controller bypasses the cache to avoid bandwidth bottlenecks, then latency is reduced, but the decision to bypass is made slowly and inaccurately by existing firmware
Solution Approach 1:
The patent replaces the complex firmware-based decision-making mechanism with a simpler processor-based moving average calculation system. This substitution enables faster and more accurate determination of when to bypass the cache, improving both request transmission speed and decision efficiency by using straightforward mathematical computations rather than complex firmware logic.
3Productivity
If the cache bus bandwidth is insufficient, then requests should be sent directly to the storage device, but existing firmware cannot reliably detect this condition
Solution Approach 1:
The patent implements a feedback mechanism by continuously measuring throughput and comparing it against a threshold derived from the cache bus bandwidth. This feedback loop enables precise detection of bandwidth saturation conditions, allowing the system to accurately determine when to bypass the cache and maximize system throughput.
Data Source
AI summary
Embodiments herein relate to sending a request to a storage device based on a moving average. A threshold is determined based on a storage device type and a bandwidth of a cache bus connecting a cache to a controller. The moving average of throughput is measured between the storage device and a host. The request of the host to access the storage device is sent directly to the storage device, if the moving average is equal to the threshold.