Write-Only Memory Page Swapping Under Memory Pressure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional memory management techniques for computing environments, such as memory swapping and kernel same-page merging, incur high I/O and CPU costs, result in inefficient memory usage, and cause undesirable latency due to frequent allocation and de-allocation of memory pages.
Innovation Solution
Implementing a write-only swapping mechanism where existing memory pages are discarded without being written to secondary storage, and new memory pages are allocated in their place, using least recently used (LRU) and out-of-memory (OOM) policies to select pages for discard, reducing CPU overhead and improving memory efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional memory swapping is performed to store and retrieve memory pages from secondary storage, then memory capacity is improved (processes can exceed physical memory limits), but I/O costs and latency increase
Solution Approach 1:
The patent extracts the essential function of memory management (allocating and deallocating pages) while removing the harmful I/O operations to secondary storage. By discarding pages in memory without swapping to disk, the system maintains memory capacity management while eliminating the time-consuming read/write operations to secondary storage devices.
Solution Approach 2:
Instead of the conventional approach of swapping pages out to secondary storage and back when needed, the patent inverts the approach by discarding pages immediately when memory pressure occurs. This reversal eliminates the round-trip I/O latency while still managing memory capacity through selective page disposal.
2Quantity of substance
If conventional memory swapping is performed to manage memory pages, then memory capacity is improved, but CPU costs increase
Solution Approach 1:
The patent removes the CPU-intensive I/O management operations from the memory swapping process. By eliminating the need to write pages to and read pages from secondary storage, the system significantly reduces CPU cycle consumption while maintaining the ability to manage memory capacity through in-memory page tracking and selective discarding.
3Adaptability or versatility
If memory pages are frequently allocated and de-allocated in conventional swapping, then memory flexibility is improved, but system performance deteriorates due to overhead
Solution Approach 1:
The patent treats memory pages as disposable objects that can be quickly allocated and discarded without expensive recovery operations. By implementing a simple discard mechanism rather than complex swap-in/swap-out operations, the system achieves high memory flexibility with minimal overhead, allowing rapid page lifecycle management that maintains system performance.
Data Source
AI summary
Techniques described herein relate to performing write-only swapping of memory pages in a computing environment. For example, the computing environment can execute one or more processes accessing memory pages. The computing environment can request, by a process of the one or more processes, a new memory page. The computing environment can determine that a memory consumption limit predefined for the memory device has been exceeded by the one or more processes. The computing environment can, in response to (i) requesting the new memory page and (ii) determining that the memory consumption limit has been exceeded, discard an existing memory page. The computing environment can allocate the new memory page for the process subsequent to discarding the existing memory page.