Low-overhead dynamic sharing of graphics memory space in GPU virtualization environments

2019 ◽  
Vol 23 (3) ◽  
pp. 2167-2178
Author(s):  
Minwoo Gu ◽  
Younghun Park ◽  
Youngjae Kim ◽  
Sungyong Park
Keyword(s):  
Memory Space ◽  
Gpu Virtualization

scholarly journals Ballooning Graphics Memory Space in Full GPU Virtualization Environments

2019 ◽  
Vol 2019 ◽  
pp. 1-11
Author(s):  
Younghun Park ◽  
Minwoo Gu ◽  
Sungyong Park
Keyword(s):  
High Performance ◽  
Virtual Machines ◽  
Graphics Processing Unit ◽  
Performance Degradation ◽  
Processing Unit ◽  
Memory Space ◽  
Memory Size ◽  
Memory Sharing ◽  
Gpu Virtualization ◽  
Graphics Processing

Advances in virtualization technology have enabled multiple virtual machines (VMs) to share resources in a physical machine (PM). With the widespread use of graphics-intensive applications, such as two-dimensional (2D) or 3D rendering, many graphics processing unit (GPU) virtualization solutions have been proposed to provide high-performance GPU services in a virtualized environment. Although elasticity is one of the major benefits in this environment, the allocation of GPU memory is still static in the sense that after the GPU memory is allocated to a VM, it is not possible to change the memory size at runtime. This causes underutilization of GPU memory or performance degradation of a GPU application due to the lack of GPU memory when an application requires a large amount of GPU memory. In this paper, we propose a GPU memory ballooning solution called gBalloon that dynamically adjusts the GPU memory size at runtime according to the GPU memory requirement of each VM and the GPU memory sharing overhead. The gBalloon extends the GPU memory size of a VM by detecting performance degradation due to the lack of GPU memory. The gBalloon also reduces the GPU memory size when the overcommitted or underutilized GPU memory of a VM creates additional overhead for the GPU context switch or the CPU load due to GPU memory sharing among the VMs. We implemented the gBalloon by modifying the gVirt, a full GPU virtualization solution for Intel’s integrated GPUs. Benchmarking results show that the gBalloon dynamically adjusts the GPU memory size at runtime, which improves the performance by up to 8% against the gVirt with 384 MB of high global graphics memory and 32% against the gVirt with 1024 MB of high global graphics memory.

scholarly journals Scalable GPU Virtualization with Dynamic Sharing of Graphics Memory Space

2018 ◽  
Vol 29 (8) ◽  
pp. 1823-1836 ◽  
Author(s):  
Mochi Xue ◽  
Jiacheng Ma ◽  
Wentai Li ◽  
Kun Tian ◽  
Yaozu Dong ◽  
...  
Keyword(s):  
Memory Space ◽  
Gpu Virtualization

A survey on cloud computing privacy and security compliances in health domain (Preprint)

2020 ◽  
Author(s):  
Zahra Zandesh
Keyword(s):  
Cloud Computing ◽  
New Paradigm ◽  
Data Governance ◽  
Healthcare Organizations ◽  
Privacy And Security ◽  
Memory Space ◽  
Health Domain ◽  
Security Services ◽  
Cloud Compliance ◽  
Complicated Nature

BACKGROUND The complicated nature of cloud computing encompassing internet-based technologies and service models for delivering IT applications, processing capability, storage, and memory space and some notable features motivate organizations to migrate their core businesses to the cloud. Consequently, healthcare organizations are much interested to migrate to this new paradigm despite challenges about security, privacy and compliances issues. OBJECTIVE The present study was conducted to investigate all related cloud compliances in health domain in order to find gaps in this context. METHODS All works on cloud compliance issues were surveyed after 2013 in health domain in PubMed, Scopus, Web of Science, and IEEE Digital Library databases. RESULTS Totally, 36 compliances had been found in this domain used in different countries for a variety of purposes. Initially, all founded compliances were divided into three parts as well as five standards, twenty-eight legislations and three policies and guidelines each of which is presented here by in detail. CONCLUSIONS Then, some main headlines like compliance management, data management, data governance, information security services, medical ethics, and patients' rights were recommended in terms of any compliance or frameworks and their corresponding patterns which should be involved in this domain.

AN EXTENDED FUZZY CLUSTERING ALGORITHM AND ITS APPLICATION

1995 ◽  
Vol 05 (02) ◽  
pp. 239-259
Author(s):  
SU HWAN KIM ◽  
SEON WOOK KIM ◽  
TAE WON RHEE
Keyword(s):  
Fuzzy Clustering ◽  
Clustering Algorithm ◽  
Clustering Algorithms ◽  
Main Memory ◽  
Color Image Segmentation ◽  
Occurrence Rate ◽  
Secondary Memory ◽  
Worst Case ◽  
Memory Space ◽  
Fuzzy Clustering Algorithm

For data analyses, it is very important to combine data with similar attribute values into a categorically homogeneous subset, called a cluster, and this technique is called clustering. Generally crisp clustering algorithms are weak in noise, because each datum should be assigned to exactly one cluster. In order to solve the problem, a fuzzy c-means, a fuzzy maximum likelihood estimation, and an optimal fuzzy clustering algorithms in the fuzzy set theory have been proposed. They, however, require a lot of processing time because of exhaustive iteration with an amount of data and their memberships. Especially large memory space results in the degradation of performance in real-time processing applications, because it takes too much time to swap between the main memory and the secondary memory. To overcome these limitations, an extended fuzzy clustering algorithm based on an unsupervised optimal fuzzy clustering algorithm is proposed in this paper. This algorithm assigns a weight factor to each distinct datum considering its occurrence rate. Also, the proposed extended fuzzy clustering algorithm considers the degree of importances of each attribute, which determines the characteristics of the data. The worst case is that the whole data has an uniformly normal distribution, which means the importance of all attributes are the same. The proposed extended fuzzy clustering algorithm has better performance than the unsupervised optimal fuzzy clustering algorithm in terms of memory space and execution time in most cases. For simulation the proposed algorithm is applied to color image segmentation. Also automatic target detection and multipeak detection are considered as applications. These schemes can be applied to any other fuzzy clustering algorithms.

Cooperative Mechanism of Local Memory and Cache in Network Processors

2013 ◽  
Vol 380-384 ◽  
pp. 1969-1972
Author(s):  
Bo Yuan ◽  
Jin Dou Fan ◽  
Bin Liu
Keyword(s):  
Packet Processing ◽  
Network Processors ◽  
Local Memory ◽  
Memory Space ◽  
Memory Mechanism ◽  
Hierarchical Memory ◽  
Packet Data ◽  
On Chip ◽  
Cache Mechanism ◽  
Cache Miss

Traditional network processors (NPs) adopt either local memory mechanism or cache mechanism as the hierarchical memory structure. The local memory mechanism usually has small on-chip memory space which is not fit for the various complicated applications. The cache mechanism is better at dealing with the temporary data which need to be read and written frequently. But in deep packet processing, cache miss occurs when reading each segment of packet. We propose a cooperative mechanism of local memory and cache. In which the packet data and temporary data are stored into local memory and cache respectively. The analysis and experimental evaluation shows that the cooperative mechanism can improve the performance of network processors and reduce processing latency with little extra resources cost.

Sign in / Sign up

Export Citation Format

Share Document