Memory Page Migration Using Sampled Access Temperature Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computer systems, processors face challenges in accurately determining access frequencies of memory pages due to the need for a large quantity of counters, especially for small pages, leading to inefficiencies in data processing speed and delay, as frequently accessed data may be stored in low-speed memories and infrequently accessed data in high-speed memories.
Innovation Solution
A method for migrating memory data involves sampling memory channels based on a sampling strategy that considers both spatial and temporal dimensions to determine access frequencies of pages, allowing for accurate identification of access temperatures, thereby triggering data migration to optimize storage based on access frequencies, with hot data moved to high-speed memories and cold data to low-speed memories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the processor accesses memories in an interleaving manner with a large quantity of counters to collect statistics on access frequencies of pages, then the accuracy of access frequency statistics is improved, but the device complexity and data processing delay increase significantly
Solution Approach 1:
The patent extracts only the necessary statistics information from the memory access stream by using sampling technology. Instead of collecting complete access frequency data for all pages, the system samples a subset of memory channels and pages to obtain representative access patterns, thereby reducing the quantity of counters and statistical data while maintaining measurement accuracy.
Solution Approach 2:
The patent applies partial action by collecting statistics on only a partial set of memory channels and pages rather than all of them. The sampling strategy selectively monitors a representative subset of memory access patterns, which is sufficient to determine access temperatures without requiring comprehensive counting of all pages, thus reducing device complexity.
2Measurement precision
If the processor collects comprehensive access frequency statistics for all pages, then the accuracy of data placement decisions is improved, but the data processing speed decreases due to the large amount of statistical data
Solution Approach 1:
The patent extracts only the essential statistics needed for access temperature determination by sampling memory channels and pages. This selective extraction maintains the accuracy required for data placement decisions while significantly reducing the volume of statistical data that must be processed, thereby improving data processing speed.
Solution Approach 2:
The system performs partial statistical collection by monitoring only a sampled subset of memory channels and pages. This partial action provides sufficient accuracy for determining access temperatures and making data placement decisions without the computational burden of processing complete access frequency data for all pages.
3Productivity
If the system stores frequently accessed data in high-speed memories and infrequently accessed data in low-speed memories, then the data processing speed is improved, but the device complexity increases due to the need for accurate access temperature determination
Solution Approach 1:
The patent extracts sufficient statistics from sampled memory channels and pages to accurately determine access temperatures. This selective extraction provides the necessary information for intelligent data placement between high-speed and low-speed memories without requiring complex comprehensive analysis of all memory access patterns, thus managing device complexity.
Solution Approach 2:
The system uses partial statistical collection through sampling to determine access temperatures. By analyzing only a representative subset of memory access patterns, the system achieves accurate temperature determination for data placement decisions without the computational complexity of processing all access data, enabling efficient data distribution across different memory types.
Data Source
AI summary
A method, implemented by one or more processors in a computer device, for migrating memory data, the method includes accessing memory media in an interleaving manner, sampling a plurality of memory channels based on a sampling strategy; determining, based on access frequencies of interleaved data blocks in sampled memory channels, access frequencies of pages in which the interleaved data blocks are located in the memory media, to determine access temperatures of the pages based on the access frequencies of the pages; and triggering data migration based on the access temperatures. All access frequencies of the pages are determined based on some access frequencies of the pages.