Performance estimation and scheduling of parallel computing programs in virtualized clusters가상 클러스터에서 실행되는 병렬 컴퓨팅 프로그램의 성능 예측 및 배치 연구

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 560
  • Download : 0
With the advancement of cloud computing, there has been a growing interest in exploiting demand-based cloud resources for parallel scientific applications. To satisfy different needs for computing resources, cloud providers provide many different types of virtual machines (VMs) with various numbers of computing cores and amounts of memory. The cost and execution time of a scientific application vary depending on the types of VMs, number of VMs, and current status of the cloud due to interference among VMs. However, currently, cloud users are solely responsible for selecting the most effective VM configuration for their needs, but often end up with sub-optimal selections. In this dissertation, using molecular dynamics simulations as a case study, we propose a framework to guide users to select the optimal VM configurations that satisfy their equirements for scientific parallel computing in virtualized clusters. For molecular dynamics computation on a cluster of VMs, the guidance framework uses artificial neural networks which are trained to predict its execution times for various inputs, VM configurations, and status of interference among VMs. Using our performance prediction mechanisms, the guidance framework helps users choose an optimal or near-optimal VM cluster configuration under cost and runtime constraints. However, estimating the execution time with status of interference does not guarantee the optimal utilization of entire cloud. Each application have different pattern of shared resource and it causes the performance variabilty of co-running application. Despite many prior studies on interference in single-node systems, the interference behaviors of distributed applications have not been investigated thoroughly. In distributed parallel applications, a local interference in a node can affect the whole execution of the application spanning many nodes. This dissertation studies an interference modeling methodology for distributed applications to predict the performance under interference in consolidated clusters. We characterize the effects of interference for various distributed applications over different interference settings, and analyzes how diverse interference intensities on the multiple nodes affect the performance. Using the proposed method, we develop a profile-based interference-aware placement algorithm based on simulated annealing to deal with two purposes, efficiently consolidate multiple distributed applications and QoS-Aware placement. Even though, profile-based interference-aware placement has a limitation due to its internal characteristic. First, profiling stage is needed before the new application scheduled. Second, the same kind of co-runner must be executed until the target workload ends. Third, several workload combination cannot be co-located due to the internal existance of interference for guaranting the high QoS. To solve limitation of profile-based study, we propose an on-line based interference aware performance estimation. By estimating the performance dynamically, the profile stage is needless, performance estimation is available even though the co-runner changes, and guaranting the target QoS for every workload combination is available by adjusting the cpu utilzation of co-runner.
Advisors
Huh, Jaehyukresearcher허재혁researcher
Description
한국과학기술원 :전산학부,
Publisher
한국과학기술원
Issue Date
2016
Identifier
325007
Language
eng
Description

학위논문(박사) - 한국과학기술원 : 전산학부, 2016.2 ,[vii, 77 p. :]

Keywords

Cloud Computing; Scientific Applications; Virtual Cluster configuration; performance modeling and prediction; interference; 클라우드 컴퓨팅; 과학응용프로그램; 가상 클러스터 설정; 성능 모형 및 예측; 간섭 현상

URI
http://hdl.handle.net/10203/222425
Link
http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=648287&flag=dissertation
Appears in Collection
CS-Theses_Ph.D.(박사논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0