Metadata Cache for Distributed Storage I/O Bottlenecks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional storage systems experience inefficiencies due to multiple context switches and bandwidth reduction when non-owning storage processors forward all requests to the owning processor, leading to CPU saturation and reduced performance.
Innovation Solution
Implementing a method where metadata objects of the common block file system are stored in a tree structure, with a cache on each storage processor, allowing local access and reducing the need for forwarding requests to the owning processor by using a common messaging interface for communication and caching metadata objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If non-owning storage processor forwards all requests to owning storage processor, then file system object access is enabled, but multiple context switches occur and CPU becomes saturated
Solution Approach 1:
The patent extracts metadata from the owning storage processor and places it in a distributed cache on non-owning storage processors. This allows non-owning processors to service read requests locally without forwarding to the owning processor, eliminating context switches and reducing CPU saturation while maintaining file system object access capability.
Solution Approach 2:
The patent performs preliminary action by pre-caching metadata on non-owning storage processors before requests arrive. This allows immediate local servicing of requests without needing to forward them to the owning processor, thereby preventing context switches and CPU saturation.
2Ease of operation
If non-owning storage processor forwards all requests to owning storage processor, then file system object access is enabled, but bandwidth on storage processors is reduced
Solution Approach 1:
The patent extracts metadata from the owning storage processor and distributes it to non-owning processors' caches. This enables local request servicing without bandwidth-consuming remote communications, while preserving the ability to access file system objects through the cached metadata.
Solution Approach 2:
The patent introduces a metadata cache as an intermediary between non-owning storage processors and the owning storage processor. This intermediary allows local servicing of requests, eliminating the need for bandwidth-intensive communication with the owning processor while maintaining file system object access.
3Productivity
If metadata is cached on non-owning storage processor, then local request servicing is enabled, but metadata consistency must be maintained
Solution Approach 1:
The patent implements feedback mechanisms where non-owning storage processors monitor metadata changes on the owning processor and update their local caches accordingly. This feedback loop maintains metadata consistency across the distributed system while enabling efficient local request servicing.
Solution Approach 2:
The patent employs periodic synchronization actions where non-owning storage processors periodically check for metadata updates from the owning processor and refresh their local caches. This periodic action maintains metadata consistency while allowing local servicing between synchronization points.
Data Source
AI summary
A storage processor having access to published logical unit numbers of a common block file system mounted on another storage processor processes requests for a particular file system object of the common block file system using metadata objects of the common block file system cached in both storage processors to access storage units shared between the two storage processors.