Optimization of out of memory killer for embedded Linux environments

Abstract

When a swapless Linux system runs out of memory, the system reclaims memory by invoking Out of Memory (OOM) Killer. The OOM killer terminates arbitrary processes. The running time of selecting victim processes is proportional to the number of processes in the system. In this paper, we propose a novel selection scheme which runs in O(1) time. The experimental results showed that the proposed selection scheme outperformed the existing one in terms of the running time. General Terms Algorithms KeywordsMemory management, Out of memory, OOM killer

From virtualization security issues to cloud protection opportunities: An in-depth analysis of system virtualization models

“…The OS kernel is in charge of system management, by providing routines that are used by application processes. Nevertheless, the kernel is also able to manage this address space for various purposes, such as process swapping, free space reclamation [31], and exception handling. All virtualization models are exposed to this vulnerability.…”

Personal Behavior-based Dynamic Governor Switching on Android System

“…Other moments of SFD monitoring are application loading and termination time. For implementation with minimum modifying of Android codes, we chained SFD monitoring code after low memory killer (LMK) of Android system [7] [8].…”

Virtual memory partitioning for enhancing application performance in mobile platforms

“…The page reclamation mechanism [10]- [12] is useful to obtain the available memory space on the system. However, it finds target pages that are sacrificial in memory reclamation, based on the least recently used page (LRU) replacement algorithm [6], [13]. It blindly handles all processes without the platform level semantics, which are important system applications in a mobile platform.…”

“…This section compares the experimental result of the number of execution frequencies of LMK/OOMK between the existing system (before) and proposed system (after). The proposed system dramatically removed the number of execution frequencies of naive memory reclamation [4]- [6] running the LMK operation and OOMK operation sequentially. The experimental result shows that untrusted applications do not increase the number of LMK operations by isolating the page reclamation of the untrusted applications executed only in the VNODE1 of the 1.5 GB memory node.…”

“…The frequent operations of LMK and out-ofmemory killer (OOMK) could seriously deteriorate userperceived performance in two ways. First, because all relevant memory space of a victim application [6] is unloaded, the unloaded memory should be reloaded at the next launching of the victim application, and it could seriously slow down the application performance. To select a victim application, OS considers the following criteria: the number of threads, the central processing unit (CPU) running time, the scheduling priority, and whether or not it directly accesses the hardware.…”