Page Allocation in Many-to-One Virtualization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Many-to-one virtualization systems face performance degradation due to high access latency caused by network-based consistency maintenance in Distributed Shared Memory (DSM) systems, leading to lower performance compared to single high-performance systems.
Innovation Solution
Optimizing the DSM system by determining whether a page fault interrupt is caused by page initialization, sending an ownership change message to the node with page ownership, and initializing the page upon receiving an ownership-change-processing-complete message, thereby minimizing processing time and optimizing performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a DSM system is used to maintain memory consistency across multiple nodes, then memory consistency is achieved, but access latency increases significantly due to network-based communication
Solution Approach 1:
The system performs preliminary actions by detecting page fault interrupts caused by page initialization and proactively sending ownership change messages before actual data transfer is needed. This prevents unnecessary network traffic and reduces access latency while maintaining memory consistency across nodes.
Solution Approach 2:
The DSM system enables nodes to self-manage page ownership and initialization states. When a node detects a page fault during initialization, it automatically handles the ownership change and page creation without requiring external intervention, reducing network overhead and improving access speed.
2Reliability
If traditional DSM page allocation is used, then memory consistency is maintained, but processing time increases due to sequential ownership change and page initialization
Solution Approach 1:
The system eliminates idle waiting time by allowing the requesting node to immediately initialize the page after sending the ownership change message, without waiting for confirmation. This continuous action approach maintains page consistency while significantly reducing processing time for memory allocation operations.
Solution Approach 2:
The ownership change message is sent in advance before the actual page data transfer is complete. The requesting node performs initialization operations preliminarily, overlapping communication and computation phases to reduce total processing time while ensuring consistency through the message protocol.
3Adaptability or versatility
If multiple nodes access shared memory pages, then system scalability is improved, but network bandwidth consumption increases due to frequent page transfers
Solution Approach 1:
The system sends ownership change messages preliminarily to establish page ownership before actual data access occurs. This prevents unnecessary page transfers across the network and reduces bandwidth consumption while maintaining the scalability benefits of multi-node access.
Solution Approach 2:
Each node autonomously manages its local page tables and handles page faults locally when possible, reducing the need for network-based page transfers. This self-service approach maintains system scalability while minimizing network bandwidth consumption by eliminating redundant data transfers.
Data Source
AI summary
Disclosed herein are an apparatus and method for page allocation in a many-to-one virtualization environment. The method may include determining whether a page fault interrupt is caused by page initialization for page allocation, sending an ownership change message to a node having ownership of the corresponding page when the page fault interrupt is determined to be caused by page initialization, and initializing the corresponding page upon receiving an ownership-change-processing-complete message.