Server Cache Data Arrangement Optimization via Dummy I/O Requests
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cache control algorithms fail to optimize data arrangement in flash memory based on application access characteristics and service schedules, leading to suboptimal performance, especially in applications that require frequent access to management information.
Innovation Solution
A computer system comprising a service server, a storage server, and a management server, where the management server analyzes access characteristics and generates control policies to optimize data arrangement in the server cache through dummy I/O requests, independent of the cache control algorithm, thereby improving processing performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a general cache control algorithm (e.g., LRU) is used to manage flash memory cache, then fast data access is achieved, but the service performance as a whole cannot be increased because the algorithm cannot control data while considering application access characteristics and service schedules
Solution Approach 1:
The patent introduces a cache driver as an intermediary layer between the cache control algorithm and the application. This cache driver intercepts I/O requests, analyzes access characteristics, and issues dummy I/O requests to manipulate cache data arrangement. The intermediary enables the system to consider application-specific access patterns while maintaining the general cache control algorithm's fast access capability.
Solution Approach 2:
The cache driver performs preliminary actions by issuing dummy I/O requests before actual data access occurs. These dummy requests pre-load or pre-position data in the cache according to predicted access patterns, ensuring that frequently accessed data is already in the cache when needed, thus improving service performance without sacrificing access speed.
2Productivity
If management information is stored in flash memory for database scanning applications, then data can be read from the database, but the general cache control algorithm ejects the management information from cache memory, requiring re-reading each time data is read
Solution Approach 1:
The cache driver implements feedback by monitoring actual I/O requests from the application and using this information to issue subsequent dummy I/O requests. The system observes that management information is repeatedly accessed, and in response, the cache driver keeps this information in the cache by issuing dummy read requests, preventing its ejection and eliminating the need for re-reading.
Solution Approach 2:
The patent changes the cache management parameters dynamically based on access characteristics. Instead of using a fixed LRU policy, the system adjusts cache retention behavior by issuing dummy I/O requests that modify the effective TTL or priority of specific cache entries, ensuring that management information remains in cache longer than standard algorithms would allow.
Data Source
AI summary
A computer system, comprising: a service server on which an application; a storage server for storing data used by the application; and a management server for managing the service server and the storage server, the service server including a cache device which a server cache for temporarily storing data is set, and including an operating system and an I/O request monitoring unit for monitoring an I/O request issued by the application, and issuing a dummy I/O request for controlling an arrangement of data in the server cache in a case where a predetermined condition is satisfied, the management server including a cache optimization unit for generating a control policy for issuing the dummy I/O request.