Download presentation
Presentation is loading. Please wait.
Published byEric Hopkins Modified over 9 years ago
2
Naixue Xiong @ GSU Slide 1 ICVCI’09 Oct. 22, 2009 A Multi-Cloud Computing Scheme for Sharing Computing Resources to Satisfy Local Cloud User Requirements N. Xiong, A. Vandenberg, M.L. Russell and K.P. Robinson Information Systems and Technology, GSU
3
Naixue Xiong @ GSU Slide 2 ICVCI’09 Oct. 22, 2009 Outline 1.Problem and Background 2.Contributions 3.VCL system model 4.VCL control scheme 5.Simulation Results 6.Conclusions and Future work
4
Naixue Xiong @ GSU Slide 3 ICVCI’09 Oct. 22, 2009 Problem How to satisfy large numbers of users applications by using limited resources in dynamically scalable, virtualized services across networks? Virtual computing is becoming increasingly important as a solution to solve it.
5
Naixue Xiong @ GSU Slide 4 ICVCI’09 Oct. 22, 2009 Background Virtual Computing Lab (VCL): An Apache.org incubator, open source code project, A remote access service A management framework Characteristics of three main parameters: processing, storage, and memory
6
Naixue Xiong @ GSU Slide 5 ICVCI’09 Oct. 22, 2009 Background VCL makes available extensive application environments: multiple applications are deployed in multiple environments; Instructors requiring course allocations reserve computing resources for future times or as allocations [1];
7
Naixue Xiong @ GSU Slide 6 ICVCI’09 Oct. 22, 2009 Background GSU is deploying VC as a solution alternative to traditional student computing labs VC as a solution to support researchers: where researchers request computing environments that may be non-standard configurations not readily available Some VCL related areas of interest are: Network control and security; dynamic virtual local area networks (VLANS) and VLAN control; support for high- performance computing (HPC); resource allocation between HPC and other services.
8
Naixue Xiong @ GSU Slide 7 ICVCI’09 Oct. 22, 2009 Outline 1.Problem and Background 2.Contributions 3.VCL system model 4.VCL control scheme 5.Simulation Results 6.Conclusions and Future work
9
Naixue Xiong @ GSU Slide 8 ICVCI’09 Oct. 22, 2009 Contributions Considers target resource utilization and the history of application computing rates Focus on dynamically scalable, virtualized services across networks to enable a multi-cloud computing scheme for sharing computing resources to satisfy local cloud user requirements. Fast response and high utilization are sometimes conflicting requirements when you have multiple users. A feedback control scheme: set and monitor both user desired response levels and target levels of resource utilizations.
10
Naixue Xiong @ GSU Slide 9 ICVCI’09 Oct. 22, 2009 Contributions propose a system model for a multi-cloud environment propose an proportional and integral scheme to satisfy requirements of multiple users. VCL scheme: considers actual versus target value of resource utilization and the history of application computing rates provide a theoretical analysis of system stability and give guidelines for selection of feedback control parameters Simulations
11
Naixue Xiong @ GSU Slide 10 ICVCI’09 Oct. 22, 2009 Outline 1.Problem and Background 2.Contributions 3.VCL system model 4.VCL control scheme 5.Simulation Results 6.Conclusions and Future work
12
Naixue Xiong @ GSU Slide 11 ICVCI’09 Oct. 22, 2009 Model North Carolina State University VCL model [1].
13
Naixue Xiong @ GSU Slide 12 ICVCI’09 Oct. 22, 2009 Model VCL Software/ Management nodes Servers Delay t 2 Delay t 1 User/applications Model: Applications sent to a management node that controls many servers’ resources (x: used; M: assigned; u: released resource)
14
Naixue Xiong @ GSU Slide 13 ICVCI’09 Oct. 22, 2009 Outline 1.Problem and Background 2.Contributions 3.VCL system model 4.VCL control scheme 5.Simulation Results 6.Conclusions and Future work
15
Naixue Xiong @ GSU Slide 14 ICVCI’09 Oct. 22, 2009 VCL scheme VCL Scheme (x: used; M: assigned; c: capacity resource) error signal between the used CPU resource and the target value for the used CPU resource sum of the history assigned CPU resource in management node during the last round-trip time used CPU resource at a server can be stabilized near its target value.
16
Naixue Xiong @ GSU Slide 15 ICVCI’09 Oct. 22, 2009 VCL scheme A recursive digital filter for VCL feedback controller: dynamic, stabilize itself based on the internal feedbacks parameters integral component proportional component + the initial incoming resource
17
Naixue Xiong @ GSU Slide 16 ICVCI’09 Oct. 22, 2009 VCL scheme apply z-transformation to Eqs. (3-4) and relevant control theory If is close to the target value, the system get better performance.
18
Naixue Xiong @ GSU Slide 17 ICVCI’09 Oct. 22, 2009 Outline 1.Problem and Background 2.Contributions 3.VCL system model 4.VCL control scheme 5.Simulation Results 6.Conclusions and Future work
19
Naixue Xiong @ GSU Slide 18 ICVCI’09 Oct. 22, 2009 Simulation Model Model: Applications sent to a management node that controls many servers’ resources x: used)
20
Naixue Xiong @ GSU Slide 19 ICVCI’09 Oct. 22, 2009 Parameters setting The target value for the servers is 80%...
21
Naixue Xiong @ GSU Slide 20 ICVCI’09 Oct. 22, 2009 Two schemes VCL controller Without a VCL controller: Application computing amount that the Management node assigns to every server is same Server works independently and the management evenly distributes all work.
22
Naixue Xiong @ GSU Slide 21 ICVCI’09 Oct. 22, 2009 Performance Evaluation Fig. 5. The CPU utilization for Server 1. Fig. 6. The CPU utilization for Server 2.
23
Naixue Xiong @ GSU Slide 22 ICVCI’09 Oct. 22, 2009 Performance Evaluation Fig. 7. The CPU utilization for Server 3. Fig. 8. computing amount in assigning the applications to servers
24
Naixue Xiong @ GSU Slide 23 ICVCI’09 Oct. 22, 2009 Without VCL controller Server 1: directly to the full utilization rate, means the load is too much for Server 1, and it cannot deal with it effectively. Server 2, stable at about 5%, low, means it can deal with more applications, but no way to assign work effectively. Server 3, quickly becomes 0, means it finishes its applications load using only part of its available CPU resources, wasting lots of CPU resource.
25
Naixue Xiong @ GSU Slide 24 ICVCI’09 Oct. 22, 2009 VCL controller Although the CPU utilization rate fluctuates at first, then stable quickly, reaching the target value of 80% It more effectively using the available CPU resources of the cloud environment It is an effective multi-cloud computing controller for ensuring both fast response and high resource utilization rates.
26
Naixue Xiong @ GSU Slide 25 ICVCI’09 Oct. 22, 2009 Outline 1.Problem and Background 2.Contributions 3.VCL system model 4.VCL control scheme 5.Simulation Results 6.Conclusions and Future work
27
Naixue Xiong @ GSU Slide 26 ICVCI’09 Oct. 22, 2009 Conclusions propose a system model for a multi-cloud environment, propose an effective proportional and integral feedback control scheme based on control theory for fast response + high utilization of cloud resources. provide a theoretical analysis of the system stability and give guidelines for selection of control parameters. Simulations demonstrate: effective multi-cloud computing controller for ensuring fast response + high resource utilization.
28
Naixue Xiong @ GSU Slide 27 ICVCI’09 Oct. 22, 2009 Future work Focus on the engineering implementation of the above control scheme in a multi-cloud computing environment that is being developed by several universities as part of the VCL open source community. proposed model can be expanded to address additional parameters related to other aspects of multi-cloud computing: the level of computing security that user applications may require, or even storage availability or dependability.
29
Naixue Xiong @ GSU Slide 28 ICVCI’09 Oct. 22, 2009 Q & A Thank You!
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.