Cluster Compute Server Storage Segmentation for Data Duplication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High performance computing systems face inefficiencies in data storage due to duplication of data across multiple compute nodes, particularly with 'read-only' data like operating systems, which requires large storage volumes and unnecessary duplication.
Innovation Solution
Implementing a cluster compute server architecture with common and dedicated storage volumes, where 'read-only' data is stored in common volumes accessible by all nodes, and modifiable data is stored in dedicated volumes, with storage controllers managing access to prevent unintended system modifications across nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is stored locally at each compute node, then each node has independent access to data, but data duplication increases storage requirements
Solution Approach 1:
The patent segments storage into two distinct types: common storage volumes for read-only data accessible by all compute nodes, and dedicated storage volumes for node-specific modifiable data. This segmentation allows read-only data to be stored once and shared across multiple nodes, reducing duplication while maintaining accessibility.
Solution Approach 2:
The common storage volume serves multiple compute nodes simultaneously, providing universal access to read-only data such as operating systems and configuration files. This multi-functional use of a single storage resource eliminates the need for each node to maintain separate copies.
2Quantity of substance
If read-only data is stored in common volumes, then storage efficiency improves, but control over data modifications becomes more complex
Solution Approach 1:
The system dynamically determines whether to serve data from common or dedicated storage volumes based on the compute node's authorization status. When a node is authorized to modify data, the system transitions from reading from common storage to writing to dedicated storage, and subsequent reads are served from the dedicated volume.
Solution Approach 2:
The fabric interface device acts as an intermediary between compute nodes and storage volumes, managing the complexity of access control. It intercepts write requests, determines authorization status, and automatically redirects unauthorized writes to the appropriate dedicated storage volume, shielding compute nodes from the complexity of storage management.
3Reliability
If unauthorized write requests are blocked, then data integrity is protected, but system responsiveness may decrease
Solution Approach 1:
The system performs preliminary authorization checks during the boot process or initial connection, determining which compute nodes are authorized to modify data in common storage volumes. This pre-authentication allows the fabric interface device to make rapid decisions during operational write requests without repeated authorization checks.
Solution Approach 2:
For authorized compute nodes, the system skips the authorization check entirely and directly processes write requests to dedicated storage volumes. This bypasses the security verification step for trusted nodes, maintaining system responsiveness while preserving security for unauthorized nodes.
Data Source
AI summary
A cluster compute server stores different types of data at different storage volumes in order to reduce data duplication at the storage volumes. The storage volumes are categorized into two classes: common storage volumes and dedicated storage volumes, wherein the common storage volumes store data to be accessed and used by multiple compute nodes (or multiple virtual servers) of the cluster compute server. The dedicated storage volumes, in contrast, store data to be accessed only by a corresponding compute node (or virtual server).