I/O Classification in Distributed Cloud Storage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current distributed cloud storage systems face challenges in providing optimal input/output (I/O) quality of service due to varying quality of service parameters such as caching, workload fingerprinting, and runtime, which can be improved for better performance.
Innovation Solution
The implementation of an apparatus that scans virtual disk images to determine partitions, identifies file systems, and generates I/O classification information, allowing for improved I/O caching, prefetching, and workload fingerprinting without requiring specific kernels or drivers from tenants, leveraging the scale-out nature of distributed storage systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If I/O quality of service parameters such as caching, workload fingerprinting, and runtime are varied to improve performance, then I/O quality of service is improved, but system complexity increases
Solution Approach 1:
The patent introduces an I/O classification component as an intermediary layer between the storage system and tenants. This component analyzes I/O operations, generates workload fingerprints, and provides classification information without requiring tenants to install or configure any software. The intermediary handles the complexity of I/O optimization internally while presenting a simple interface to tenants, thus improving I/O QoS without increasing tenant-side system complexity.
Solution Approach 2:
The storage system performs self-service by automatically analyzing I/O patterns, generating workload fingerprints, and optimizing caching and prefetching operations without external intervention. The system autonomously classifies I/O operations and adjusts performance parameters based on observed workloads, eliminating the need for tenants to manually configure complex QoS parameters while still achieving performance improvement.
2Productivity
If I/O classification information is generated through scanning virtual disk images, then I/O performance is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary scanning and classification of virtual disk images during system initialization or offline periods, before actual I/O operations begin. By pre-generating workload fingerprints and classification information in advance, the system avoids time-consuming analysis during runtime, thus improving I/O performance without significantly impacting operational processing time.
Solution Approach 2:
The system implements periodic scanning and reclassification of virtual disk images at scheduled intervals rather than continuously. This periodic action allows the system to update I/O classification information efficiently without constant processing overhead, balancing performance improvement with acceptable processing time consumption.
Data Source
AI summary
Embodiments of the present disclosure are directed towards techniques and configurations for an apparatus configured to provide I/O classification information in a distributed cloud storage system, in accordance with some embodiments. In one embodiment, the apparatus may include a partition scanner, to scan an image of a virtual disk associated with the storage system, to determine one or more partitions associated with the virtual disk; a file system scanner coupled with the partition scanner, to identify file systems associated with the determined partitions, to access files stored in the identified file systems; and I/O classifier coupled with the file system scanner, to generate I/O classification information associated with the accessed files. The I/O classification information provides characteristics of input-output operations performed on the virtual disk. Other embodiments may be described and/or claimed.