RAID Strip Hash Placement for In-Drive Compute Requests
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional RAID arrays face challenges in efficiently and cost-effectively performing compute operations due to the compute-intensive process of cryptographic hash value calculations, leading to inefficient use of existing compute components and high implementation costs.
Innovation Solution
A method for RAID arrays that involves identifying strip hash values, selecting a drive to perform compute requests, reading missing hash values from other drives, and writing them into spare storage space on the selected drive, allowing compute operations to be performed while ensuring hash arrays are deduplicated.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If cryptographic hash values are calculated using existing compute components, then implementation cost is reduced, but compute throughput is significantly consumed
Solution Approach 1:
The patent segments the hash calculation workload by creating dedicated hardware components (hash calculation units) that are distributed across multiple drives in the RAID array. Each drive has its own hash calculation capability, allowing parallel processing of hash values without consuming the throughput of the main compute components. This segmentation resolves the contradiction by providing low-cost implementation through existing hardware resources while maintaining high compute throughput through distributed parallel processing.
2Productivity
If dedicated hardware is used to calculate hash values, then compute throughput is preserved, but implementation cost increases
Solution Approach 1:
The patent makes existing drive components multi-functional by enabling them to perform both data storage and hash calculation functions. The hash calculation units are integrated into the existing drive architecture, allowing the same hardware to serve dual purposes. This universality resolves the contradiction by preserving compute throughput through dedicated hash calculation capability while avoiding the need for separate dedicated hardware, thereby reducing implementation cost.
3Reliability
If hash values are distributed across multiple drives, then data availability is improved, but compute operation efficiency decreases
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing hash values in a distributed hash table across multiple drives before compute operations are needed. When a compute operation is requested, the system can quickly locate and retrieve the necessary hash values from the distributed structure without performing time-consuming calculations during the operation itself. This preliminary preparation resolves the contradiction by maintaining data availability through distribution while improving compute operation efficiency through pre-computed lookup capabilities.
Data Source
AI summary
A method, according to one approach, includes: receiving a compute request that corresponds to drive strips in a RAID array, and determining strip hash values that correspond to the respective drive strips. Drives in the RAID array that include one or more of the strip hash values are identified, and one of the identified drives is selected to conduct the compute request. The method also includes identifying one or more of the strip hash values that are not included in the selected drive. Strip data associated with the strip hash values identified as not being in the selected drive are read from other ones of the identified drives. Moreover, the strip data read from the other ones of the identified drives are written into spare storage space of the selected drive. The method further includes causing the selected drive to perform the compute request on the strip data therein.