Ephemeral Storage Management for Power Loss Data Preservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud computing systems face inefficiencies in managing computing resources, particularly in allocating hardware resources such as memory and storage, leading to suboptimal performance and increased costs due to inefficient processing and storage allocations.
Innovation Solution
Implementing an ephemeral data management system that detects power loss events and selectively stores critical data from volatile storage to non-volatile storage, reducing the demand for capacitors and allowing for more efficient allocation of resources, thereby improving processing and storage capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional resource allocation is used, then hardware resources are allocated to computing nodes, but processing resources and hardware real estate are allocated inefficiently
Solution Approach 1:
The patent segments storage resources into ephemeral storage (for temporary compute services) and persistent storage (for long-term data), allowing different allocation strategies for different service types. This segmentation enables compute nodes to have dedicated ephemeral storage without requiring full persistent storage capabilities, improving resource allocation efficiency.
Solution Approach 2:
The patent creates a universal storage management system that can dynamically allocate storage resources between compute services and storage services based on demand. The same infrastructure serves multiple functions: providing ephemeral storage for compute nodes, persistent storage for data services, and enabling efficient resource reallocation without requiring separate dedicated systems.
2Reliability
If ephemeral storage is managed without power loss detection, then system simplicity is maintained, but data loss occurs during power loss events
Solution Approach 1:
The system performs preliminary actions by detecting power loss events and initiating data preservation procedures before actual data loss can occur. The ephemeral storage management system monitors for power loss conditions and automatically triggers copy operations to persistent storage, ensuring data protection without requiring complex manual intervention.
Solution Approach 2:
The patent introduces an intermediary ephemeral storage management system that sits between the compute services and persistent storage. This intermediary detects power loss events and coordinates the preservation of ephemeral data to persistent storage, simplifying the overall system architecture while providing reliable data protection.
3Reliability
If capacitor capacity is increased to protect against power loss, then data protection improves, but hardware cost and complexity increase
Solution Approach 1:
The patent extracts the data protection function from the capacitor system and relocates it to the ephemeral storage management system. Instead of relying on large capacitors to maintain power during outages, the system uses intelligent detection and automated data copying to persistent storage, dramatically reducing the required capacitor capacity while improving reliability.
Solution Approach 2:
The system treats ephemeral storage as disposable temporary storage that can be easily reset and repurposed. Rather than investing in expensive, high-capacity capacitors for long-term data protection, the system uses low-cost ephemeral storage with automated preservation mechanisms, achieving better protection at lower hardware cost.
Data Source
AI summary
The present disclosure relates to systems, methods, and computer readable media for identifying and responding to one or more power loss events on a computing node. For example, systems disclosed herein may relate to management of a power loss event on a computing node hosting one or more compute platforms thereon. The systems disclosed herein may implement a power handling configuration that identified a subset of data from a volatile portion of a storage system to prioritize for storage on a non-volatile portion of the storage system. By selectively identifying and flushing data associated with compute platforms hosted by a computing node, systems described herein may significantly reduce demand for capacitance on cloud computing systems while optimizing other performance parameters (e.g., write performance, hardware durability) of devices on cloud computing systems.


