Memory Page Sharing Across Hosts via Hash Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtualized computing environments, existing methods are inefficient in sharing physical memory pages across multiple virtual machines running on different host computing systems, leading to redundant copies of identical memory pages, which increases memory overhead and reduces resource utilization.
Innovation Solution
A system and method that utilizes a memory page sharing module to identify and migrate workloads with identical physical memory pages across host computing systems, employing hash tables and remote direct memory access (RDMA) to synchronize and compare hash values, recommending live migration and transparent memory sharing to consolidate identical pages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If transparent memory sharing is implemented to reclaim physical memory pages with identical contents, then memory overhead is reduced, but the complexity of detecting and comparing memory pages across multiple host computing systems increases
Solution Approach 1:
The patent replaces the mechanical approach of directly comparing physical memory page contents with a computational hash-based system. Each physical memory page is assigned a hash value, and these hash values are stored in hash tables on host computing systems. The management server retrieves and compares hash values remotely using RDMA, eliminating the need for direct memory page comparison and significantly reducing the difficulty of detecting identical pages across distributed systems.
2Productivity
If hash tables are used to store and compare memory page identifiers across distributed hosts, then memory page identification efficiency is improved, but the device complexity increases due to additional data structures and synchronization mechanisms
Solution Approach 1:
The patent introduces a management server as an intermediary between host computing systems to coordinate hash table synchronization. The management server retrieves hash values from host systems, performs centralized comparison, and identifies pages eligible for sharing. This intermediary approach simplifies the complexity by providing a single point of coordination rather than requiring direct peer-to-peer synchronization between all host systems.
Solution Approach 2:
The patent uses hash values as copies or representations of the actual memory page contents. Instead of comparing the full memory page data, the system compares these compact hash value copies, which significantly improves identification efficiency while reducing the complexity of data transmission and comparison operations across the distributed system.
3Quantity of substance
If live migration is recommended for workloads with identical memory pages, then memory sharing efficiency is improved, but the loss of time for migration operations increases
Solution Approach 1:
The patent performs preliminary identification of identical memory pages across host systems before migration is needed. By continuously monitoring and comparing hash values, the system pre-identifies candidates for memory sharing, so that when live migration becomes necessary, the decision can be made quickly based on pre-analyzed data rather than performing complex comparisons during the migration process itself.
Data Source
AI summary
System and method for increasing physical memory page sharing by workloads executing on different host computing systems are described. In one embodiment, workloads executing on different host computing systems that access physical memory pages having identical contents are identified. Further, migration to consolidate the identified workloads on a single host computing system such that the physical memory pages can be shared using a page sharing mechanism is recommended.