Graph-Based Cache Prefetching for Hybrid Cloud Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Hybrid cloud architectures face high latency issues due to the non-local nature of components, which affects data access performance across network connections, particularly when accessing data stored in cloud computing systems from on-premise systems.
Innovation Solution
A method for prefetching data in a hybrid cloud system by analyzing access patterns using a graph structure to identify and prefetch data items based on their weighting scores, which are calculated from link weights and access sequences, allowing for proactive data retrieval from off-premise computing systems to reduce latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Volume of stationary object
If data is stored in off-premise cloud computing systems to extend storage capabilities, then storage capacity is improved, but data access latency increases due to network connections
Solution Approach 1:
The system performs preliminary actions by prefetching data items from off-premise storage to on-premise cache storage before they are actually requested. The machine learning model predicts which data items will be needed soon and retrieves them in advance, so that when applications request the data, it is already available locally, reducing access latency while maintaining the benefit of extended storage capacity
Solution Approach 2:
The patent introduces an on-premise cache storage system as an intermediary between applications and off-premise cloud storage. This intermediary layer stores frequently accessed or predicted data items locally, allowing applications to access data with low latency while the system maintains the ability to access vast off-premise storage capacity when needed
2Loss of time
If all data items are prefetched from off-premise storage to on-premise cache, then data access latency is reduced, but network bandwidth consumption increases
Solution Approach 1:
The system changes the parameter of data selection by using a machine learning model to dynamically determine which data items to prefetch based on predicted access patterns. Instead of prefetching all or no data, the model calculates a priority score for each data item and selectively prefetches only those with high priority, optimizing the balance between reducing latency and conserving network bandwidth
Solution Approach 2:
The patent applies local quality by treating different data items differently based on their predicted utility. The machine learning model assigns different priority levels to different data items, and the system prefetches data with high priority locally while leaving low-priority data in off-premise storage, creating a differentiated caching strategy that optimizes network resource usage
Data Source
AI summary
Prefetching techniques for a hybrid cloud system are described. The techniques include maintaining a graph in a graph database, the graph including nodes associated with data items to be accessed and links between the nodes, each link including an associated link weight. Conceptually, the link weights are indicative of how likely it is for a data item to which the link points to be accessed after the data item associated with the node from which the link extends is accessed. The link weights are updated as items are accessed and the graph is traversed. The graph structure and associated techniques help to reduce latency in a hybrid cloud system by providing for prefetch of items anticipated to be used at a later time.