Cloud Volume Provisioning Using Semantic Storage Class Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud volume provisioning methods, such as those used in Kubernetes clusters, often result in sub-optimal storage volume placement due to the lack of consideration for network resources and load balancing, leading to inefficiencies and unsatisfied application requirements.
Innovation Solution
A vector database is created to store historical performance metrics of storage devices, allowing for semantic searching to identify storage classes that best match application requirements, ensuring optimal placement of storage volumes based on performance, availability, and cost considerations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional cloud volume provisioning methods are used, then storage volumes can be provisioned quickly, but the placement is sub-optimal and does not satisfy application requirements
Solution Approach 1:
The system performs preliminary actions by building a vector database that stores historical performance metrics of storage devices before provisioning occurs. This pre-computed database enables rapid semantic searching during provisioning, allowing the system to quickly identify optimal storage classes without performing complex analysis in real-time, thus maintaining both high reliability and productivity.
Solution Approach 2:
The patent introduces a vector database as an intermediary between the provisioning request and the final storage device selection. This intermediary layer stores historical performance data and enables semantic searching, allowing the system to match application requirements with historically proven storage classes without directly analyzing raw performance data during provisioning, thereby improving both accuracy and speed.
2Reliability
If storage volumes are provisioned without considering historical performance, then provisioning is simpler and faster, but performance and availability are compromised
Solution Approach 1:
The system creates a simplified representation (copy) of historical performance data in the form of vectors within a vector database. Instead of working with complex raw performance metrics directly, the system uses these pre-processed vector representations that capture essential performance characteristics, enabling reliable storage selection without the complexity of processing raw historical data during provisioning.
Solution Approach 2:
The patent transforms complex historical performance metrics into simplified vector parameters that can be efficiently searched and compared. By changing the representation of performance data from raw metrics to vector embeddings, the system maintains the richness of performance information while reducing the computational complexity of analyzing it during the provisioning process.
3Reliability
If semantic search on vector database is performed, then optimal storage class is identified based on application requirements, but the process takes more time
Solution Approach 1:
The vector database is built in advance by storing historical performance metrics before provisioning requests arrive. This preliminary preparation allows the semantic search to query pre-organized data structures rather than performing analysis in real-time, significantly reducing provisioning time while maintaining optimal storage placement decisions.
Solution Approach 2:
The patent replaces traditional mechanical search methods (iterating through all storage devices and comparing metrics) with semantic search technology that uses vector mathematics. This substitution allows for rapid similarity searching in high-dimensional spaces, achieving both optimal placement and fast query response times.
Data Source
AI summary
Adaptive provisioning of cloud storage volumes includes building a vector database having properties of storage devices of a cloud environment, associating storage devices with storage classes based on the properties of the storage devices, each storage device of the storage devices being associated with a storage class of the storage classes, receiving a request for provisioning a storage volume to support a workload, where the request indicates application requirements associated with servicing the workload, performing a semantic search on the vector database and determining, based on the semantic search, a storage class for the requested storage volume, and provisioning the storage volume on a storage device, of the storage devices, associated with the determined storage class.