Object Storage Hashing for Tenant-Aware Autoscaling

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Cloud computing infrastructures face inefficiencies in resource utilization due to random distribution of data objects across object storage devices, leading to underutilized resources when tenants are inactive, as conventional deterministic hashing does not account for tenant-specific data organization.

Innovation Solution

Implement a multilevel hashing scheme that utilizes tenant profile information to store data objects on specific object storage devices, allowing for autoscaling based on active compute jobs, enabling standby or shutdown of underutilized devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If conventional deterministic hashing is used to distribute data objects, then data storage is simplified, but resource utilization efficiency deteriorates due to random distribution causing underutilization

Engineering Contradiction:

Improveresource utilization efficiencyVSAvoidhashing scheme complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The hashing scheme is segmented into multiple levels: a first hash function determines placement groups, and subsequent hash functions process tenant profile data to select specific storage devices within groups. This segmentation allows organized distribution while maintaining scalability.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different aspects of data are processed through different hash functions at different levels. Tenant profile data is specifically processed through subsequent hash functions to ensure related data from the same tenant is co-located, improving resource utilization for active tenants while allowing inactivity-based scaling.

Inventive Principle:

Principle #3Local quality

2Ease of operation

If data objects are randomly distributed across storage devices, then storage system simplicity is maintained, but the ability to shut down underutilized devices deteriorates

Engineering Contradiction:

Improvedevice shutdown capabilityVSAvoiddata accessibility

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The system dynamically scales storage devices based on active compute jobs associated with data objects. When no compute jobs are active for data on a storage device, the device can be shut down or scaled back. The multilevel hashing ensures that when devices are scaled down, data accessibility for active tenants is maintained through proper data organization.

Inventive Principle:

Principle #15Dynamics

3Productivity

If multilevel hashing with tenant profile data is implemented, then data organization by tenant is improved, but hashing computation time increases

Engineering Contradiction:

Improvedata organization efficiencyVSAvoidhash computation time

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The hashing computation is segmented into multiple stages: a first hash on data objects to determine placement groups, followed by subsequent hashes on tenant profile data. This segmentation allows for efficient processing by breaking down the computation into manageable steps that can be executed in sequence.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS12602364B2Scalable object storage

Publication Date: 2026.04.14 RED HAT LLC

AI summary

According to one example, a method includes receiving from a client device, a data object for storage within an object storage system, performing a plurality of hashes on the data object tenant profile data associated with the data object to determine one of a plurality of object storage devices to which to store the data object, and autoscaling the object storage device based on active compute jobs associated with data objects stored on the object storage device.