Fast Restart of Large Memory Systems via Storage Replica

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current large memory systems face challenges in quickly restarting after a software crash, particularly with large memory configurations, where the process of rebuilding memory state can be time-consuming and inefficient, leading to prolonged downtime and data loss.

Innovation Solution

The implementation of a software-defined server architecture with a dynamic adaptive reconfiguration framework and storage replica data structure allows for rapid rebooting by tracking and mapping guest physical memory to real physical memory, bypassing disk access and maintaining memory state integrity, enabling fast restarts without losing intermediate computation state.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If traditional memory systems are used with large memory configurations, then memory capacity is improved, but restart time increases significantly

Engineering Contradiction:
Improvememory capacityVSAvoidrestart time
Core Design Contradiction:
Quantity of substanceVSLoss of time

Solution Approach 1:

The patent implements preliminary action by maintaining a mapping between guest physical memory addresses and real physical memory addresses in a data structure before a crash occurs. This mapping is pre-established and stored, allowing the system to quickly reconstruct memory state after a crash without needing to reinitialize the entire memory space, thus dramatically reducing restart time while supporting large memory capacities.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses copying by creating a virtualized memory layer that copies the mapping relationship between guest physical memory and real physical memory. This virtual mapping structure allows the system to reconstruct memory state from the copied mapping information rather than reinitializing actual hardware memory, enabling fast restarts in large memory systems.

Inventive Principle:
Principle #26Copying

2Reliability

If memory state is rebuilt from scratch after crash, then data integrity is ensured, but restart duration increases

Engineering Contradiction:
Improvedata integrityVSAvoidrestart duration
Core Design Contradiction:
ReliabilityVSDuration of action of moving object

Solution Approach 1:

The system performs preliminary action by pre-establishing and storing the mapping between guest physical memory addresses and real physical memory addresses in a data structure before the crash occurs. This pre-built mapping serves as a recovery guide that ensures data integrity during restart while avoiding the need to rebuild memory state from scratch, thus reducing restart duration.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary mapping data structure that mediates between the guest virtual memory space and the physical hardware memory. This intermediary layer preserves the relationship between virtual and physical addresses, allowing the system to quickly restore memory state by following the pre-stored mappings rather than performing a complete memory initialization, thereby maintaining both reliability and speed.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Adaptability or versatility

If disk access is used for memory state recovery, then fallback option is provided, but performance deteriorates

Engineering Contradiction:
Improverecovery optionVSAvoidrecovery speed
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent applies copying by creating a virtualized memory mapping structure that copies the address translation information between guest physical memory and real physical memory. This copied mapping data is stored in a data structure that can be quickly accessed during recovery, providing a fast in-memory recovery mechanism that eliminates the need for slow disk access while maintaining adaptability for various crash scenarios.

Inventive Principle:
Principle #26Copying

Data Source

PatentUS20240152286A1Fast restart of large memory systems
Publication Date: 2024.05.09 HEWLETT PACKARD ENTERPRISE DEV LP
  • US20240152286A1 patent drawing
  • US20240152286A1 patent drawing
  • US20240152286A1 patent drawing

AI summary

Utilizing a storage replica data structure includes receiving, at a hyper-kernel running on a computing node in a plurality of interconnected computing nodes, an indication of an operation pertaining to at least one of a guest physical memory address or a stable storage address. A guest operating system is run on a virtual environment that is defined by a set of hyper-kernels running on the plurality of interconnected computing nodes. It further includes updating a storage replica data structure. The storage replica data structure comprises a set of entries. The set of entries in the storage replica data structure comprises associations among guest physical memory addresses, physical memory addresses, and stable storage addresses