Computational Storage Offloading Using Metadata and Latency Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computational storage systems fail to optimize computational storage operations by efficiently distributing them to storage devices with the most capable hardware or least overload, leading to increased latency and reduced bandwidth.
Innovation Solution
A method and system that leverage metadata and computational storage latency metrics to identify and redirect computational storage operations to storage devices with optimal hardware capabilities and lower load, using deduplication, snapshot, and replication metadata to determine offload locations and estimate latency for efficient distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If computational storage operations are handled by storage devices without efficient distribution, then device complexity is reduced, but latency increases and bandwidth decreases
Solution Approach 1:
The patent segments computational storage operations by distributing them to multiple storage devices based on their computational capabilities and current load. The storage controller divides operations into separate queues for different storage devices, allowing parallel processing and reducing latency by avoiding single-point bottlenecks.
Solution Approach 2:
The storage controller acts as an intermediary that receives computational storage operations, inspects them, and redirects them to appropriate storage devices. It uses metadata and latency metrics as intermediaries to make intelligent routing decisions, optimizing the flow of operations without requiring direct complex coordination between all devices.
2Device complexity
If computational storage operations are not distributed to devices with optimal hardware capabilities, then system simplicity is maintained, but operational efficiency decreases
Solution Approach 1:
The patent changes the routing parameters for computational storage operations based on storage device characteristics. It inspects operations and modifies their destination based on hardware capabilities, current load, and latency metrics, dynamically adjusting parameters to optimize throughput while maintaining manageable system complexity.
3Ease of operation
If storage devices are not selected based on computational capabilities, then device selection is simplified, but bandwidth utilization is reduced
Solution Approach 1:
The storage controller continuously monitors storage device performance metrics including latency and bandwidth utilization. This feedback is used to dynamically select optimal devices for computational storage operations, ensuring high bandwidth utilization while keeping the selection process automated and simple through policy-based routing.
Data Source
AI summary
A computer-implemented method, computer program product, and computer system for managing computational storage operations. The computer-implemented method includes receiving a computational storage operation at a storage controller and inspecting the computation storage operation to determine a type of operation. The computer-implemented method further includes leveraging metadata of the storage controller to identify possible offload locations on storage devices for handling the computational storage operation of the determined type. The computer-implemented method further includes inspecting a register of storage devices maintained with the computational capabilities of the storage devices to determine which of the identified storage devices can support the computational storage operation to obtain candidate storage devices. The computer-implemented method further includes comparing the candidate storage devices using an estimated computational storage latency metric to select a storage device for handling the computational storage operation.