Cooperative Multilayer Cache for Storage Referral Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current file system referral techniques introduce delays and latencies due to multiple redirection hops and lack robust error handling, impacting the performance of data access and backup operations in storage systems.
Innovation Solution
Implementing a cooperative multilayer-cache technique that includes client and server referral caches to optimize storage unit location tracking, reducing the need for database lookups and improving resiliency by synchronizing cache operations and propagating failure statistics across layers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Volume of stationary object
If multiple redirection hops are used to reach storage appliances, then storage capacity is extended, but latency and delays increase
Solution Approach 1:
The patent implements referral caching at multiple layers (client-side, server-side, database-side) to perform location lookups in advance and store the results. This preliminary action eliminates the need for repeated redirection hops during subsequent access operations, thereby reducing latency while maintaining extended storage capacity.
Solution Approach 2:
The patent introduces cache layers as intermediary components between the application and the storage appliances. These caches act as mediators that store referral information and provide fast local responses, eliminating the need for multiple network redirection hops and reducing overall latency.
2Volume of stationary object
If multiple redirection hops are used to reach storage appliances, then storage capacity is extended, but system complexity increases
Solution Approach 1:
The patent segments the referral handling system into multiple independent cache layers (client-side, server-side, database-side), each managing its own portion of the referral resolution process. This segmentation distributes complexity across manageable components while maintaining the ability to extend storage capacity through referrals.
Solution Approach 2:
The patent creates copies of referral information at multiple cache layers. Instead of relying on a single complex redirection path, multiple copies of location information are stored at different levels, simplifying the system by providing redundant, easier-to-access information at each layer.
3Measurement precision
If database lookups are performed for every storage unit location, then accurate location information is obtained, but performance deteriorates
Solution Approach 1:
The patent performs database lookups in advance and stores the results in cache layers before they are needed for actual storage access operations. This preliminary action ensures accurate location information is available while eliminating repeated database queries during subsequent operations, thereby maintaining performance.
Solution Approach 2:
The patent implements feedback mechanisms where cache hit/miss statistics are propagated across layers to optimize future lookup behavior. This feedback allows the system to learn from access patterns and reduce unnecessary database queries, maintaining both accuracy and performance.
4Reliability
If robust error handling is implemented in referral processing, then resiliency is improved, but processing time increases
Solution Approach 1:
The patent performs error handling and failure detection in advance by monitoring cache statistics and propagation feedback signals. This preliminary detection allows the system to prepare alternative paths or handle errors before they impact actual storage operations, improving resiliency without significant time penalty.
Solution Approach 2:
The patent uses feedback from cache statistics and failure propagation to dynamically adjust error handling behavior. This feedback mechanism allows robust error handling to be activated only when necessary, maintaining resiliency while minimizing processing time during normal operations.
Data Source
AI summary
A database manager maintains a database including information tracking a set of storage units and a set of storage appliances on which the storage units currently reside. At least a portion of the tracking information is cached in a server cache at a namespace redirection server. Some of the tracking information is cached in a set of client caches at a set of clients. A request to access a storage unit is received at a client. A lookup is made into a client cache at the client to find a storage appliance having the storage unit. A determination is made that there is a problem accessing the storage unit using the client cache. So, the namespace redirection server is queried for updated tracking information concerning a location of the storage unit. The client assists the namespace redirection server with management of the server cache.