Exploiting heterogeneity of communication channels for efficient GPU selection on multi-GPU nodes

Faraji, Iman; Mirsadeghi, Seyed H.; Afsahi, Ahmad

doi:10.1016/j.parco.2017.07.001

Cited by 4 publications

(1 citation statement)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…But previous algorithms, which can be modified for the distributed application, cannot reach nowadays line speed because they try to improve speed only depending on high-speed memory and ignore the potential computing complexity. Parallel computing, such as GPGPU, would be a good solution to this problem [5][6] [7] [8].…”

Section: Introductionmentioning

confidence: 99%

GPU based Real-time Super Hosts Detection at Distributed Edge Routers

2019

Preprint

View full text Add to dashboard Cite

The super host is a special host on the network which contacts with many other hosts during a certain time window. They play important roles in network researches such as scanners detection, resource allocation, spam filtering and so on. How to find super hosts in real time is the foundation of these applications. In this paper, a novel algorithm, denoted as CBAA, is proposed to solve this problem at edge routers. CBAA divides network traffic into different parts. A cube of bits array is devised to store hosts linking information of different traffic parts when scanning packets. At the end of each time window, CBAA restores super hosts very fast because there are only a fraction of super hosts in each traffic part. CBAA is also a parallel algorithm. Its easy to deploy CBAA in GPU to deal with highspeed network traffic in real time. Experiments on a real-world core network prove the advantage of our algorithm.

show abstract

Section: Introductionmentioning

confidence: 99%