MCEVE A Model for Configuring Efficient Virtualized Environment Based on Multiple Weighted Considerations Abdullah Almurayh MSCS Graduate Candidate Master Thesis Committee members: Dr. Edward Chow (Advisor) Dr. Chuan Yue Dr. Albert Glock Fall 2011
Outline Introduction to Virtualization Technology Virtual Machine/Virtual Private Server The Problem The Proposed Model Evaluation Lessons Learned Future Work Conclusion Demo & Questions 1 MCEVE / Abdullah Almurayh 8/26/2011
Introduction to Virtualization 2 MCEVE / Abdullah Almurayh 8/26/2011 Widely used technology Benefits of Virtualization Consolidation and isolation Reduced power and cooling Green computing Ease of deployment and administration High availability and disaster recovery Applications of virtualization Education Software Evaluation Enabling the dynamic data center Cloud computing Load Balancing Information Technology Departments Disaster Recovery Personal use
Introduction to Virtualization Virtualization types: Full-virtualization Para-virtualization Operating system-level virtualization Emulation Virtualization Projects: Xen VMware Windows Server 2008 R2 – Hyper-V OpenVZ Red Hat Virtualization RHEV Virtual box Many companies, datacenters, organizations, universities, and IT have virtualized their servers. Even Small business and individuals started using their virtualization solutions. Availability of Low-cost Public Clouds, e.g. Amazon AWS 3 MCEVE / Abdullah Almurayh 8/26/2011
Virtual Guest Virtual Machine (VM): Guest Operating System. More flexibility. Full Virtualization, Para- Virtualization. Xen, Vmware, Virtual box, RHEV. Virtual Private Server(VPS): Share host Operating System. Less flexibility when the host Kernel more efficient. OS-level Virtualization. Linux-Vserver, OpenVZ. 4 MCEVE / Abdullah Almurayh 8/26/2011
VM vs. VPS VM vs. VPS: Xen [ Para-Virtualization Virtual Machine] VMware [ Full Virtualization Virtual Machine] OpenVZ [ OS-level Virtualization Virtual Private Server] 5 MCEVE / Abdullah Almurayh 8/26/2011 Figure: Comparison of read and write performance by different file sizes Figure: Comparison of latency performance by different message sizes Figures are Cited from: Chaudhary, V.; Minsuk Cha; Walters, J.P.; Guercio, S.; Gallo, S.;, "A Comparison of Virtualization Technologies for HPC," Advanced Information Networking and Applications, AINA nd International Conference on, vol., no., pp , March 2008,
The Problem Diversity of virtualization solutions has opened the door to an endless array of choices. Virtualization technologies operate in slightly different manners. Virtualization technologies have different architectures and requirements. By having vast choices, people sometimes become so confused and unable to choose the right virtualization solutions. 6 MCEVE / Abdullah Almurayh 8/26/2011
Related Work Optimizing utilization of resource pools in web application servers By: Alexander Totok, Vijay Karamcheti, Concurrency and Computation: Practice and Experience, vol. 22 (2010), pp Research work in the area of modeling underlying server environments produces different results. This research work can also include bottleneck identification and tuning to identify system metrics for performance enhancement. Includes identification of different application configuration parameters to determine performance goals. The process of configuring virtualized environments is to achieve performance goals by producing better decisions of making virtualized environments. The proposed research is focused on the model of application configuration. 7 MCEVE / Abdullah Almurayh 8/26/2011
Related Work Quality in use: Meeting user needs for quality By: NigelBevan. Journal of Systems and Software. ACM. Dec,1999.Pages This paper describes a framework for software product quality developed for: Internal quality: static properties of the code. External quality: behaviour of the software when it is executed. Quality in use: the extent to which the software meets the needs of the user. The paper defines the quality in use as a broader view of different concepts such as functionality, reliability, usability, efficiency, and the like. This framework is needed for evaluating the MCEVE software or application. 8 MCEVE / Abdullah Almurayh 8/26/2011
Related Work A Quality of Service Management Framework Based on User Expectations Vikas Deora, J. Shao, W. Alex Gray and Nick J. Fiddian, Service-Oriented Computing - ICSOC 2003, p This paper presents a quality of service management framework based on user expectations by collecting expectations as well as ratings from the users of a service then calculating the quality of the service. This approach does not allow the user to specify, for example, the minimum and maximum expectations. MCEVE is also based on user expectations; however, it allows the user to specify weights of these expectations. 9 MCEVE / Abdullah Almurayh 8/26/2011
Virtualization Solutions not Equal 10 MCEVE / Abdullah Almurayh 8/26/2011 CPU scalability of different leading virtualization solutions in the UNIX and distributed server market Cited from: Not All Server Virtualization Solutions are Created Equal By: Andre Metelo IBM SWG Competitive Project Office. 08/13/2010
Trade-off Complexity of Platforms vs. Benchmarks 11 MCEVE / Abdullah Almurayh 8/26/2011 Native VMware Xen Figure: Passmark – CPU results compared to native (higher is better).. Cited From: VMware, "A Performance Comparison of Hypervisors" Native Xen VMware User-mode Figure : Relative performance of native Linux (L), XenoLinux (X), VMware workstation 3.2 (V) and User-Mode Linux (U) (higher is better). Cited From: P. Barham, B. Dragovic, K. Fraser, S. Hand, T. Harris, A. Ho, et a, "Xen and the art of virtualization," in In Proc. Of the 19th ACM Symposium on Operating System Principles, Bolton Landing, NY, Oct Relative score to native (higher is better)
The Proposed Model Develop a model for suggesting better solutions of virtualized environments based on the user weighted considerations. Use user weighted considerations as inputs for algorithmic outputs To have suggested solutions in easy way and low cost. A user can get a good overview of configurations that may meet his expectations. 12 MCEVE / Abdullah Almurayh 8/26/2011
Page 13 The Goals for the Proposed Model Cost: Adopting the appropriate decisions resulting in hours configuring virtualized environments instead of spending days researching and comparing existing results. The cost can be reduced by the use of existing reliable results instead of performing tests and experiments that cost money and time. Performance: Performance can be enhanced by using solutions that are based on the best performance comparisons. Trying different unbeknown solutions may have potential failures and lead to inefficient virtualized environments. 13 MCEVE / Abdullah Almurayh 8/26/2011
Page 14 The Proposed Model 14 MCEVE / Abdullah Almurayh 8/26/2011 Different opinions lead to different decisions combined views
Page 15 The Algorithm of the Model 15 MCEVE / Abdullah Almurayh 8/26/2011 Conf: user configuration Weight: user configuration weight Measure: benchmark measurement W: Weight value M: measurement value Mix: overall calculation n: number of resulted solution best: best selection of the overall results.
Page 16 The Flow Diagram of the Model 16 MCEVE / Abdullah Almurayh 8/26/2011
Page 17 Benchmark Data Collection SPCE virtualization measurements (SPECvirt) 17 MCEVE / Abdullah Almurayh 8/26/2011 Example
Page 18 MCEVE Implementation 18 MCEVE / Abdullah Almurayh 8/26/2011 Web Application Based Implementation
Page 19 Input of MCEVE 19 MCEVE / Abdullah Almurayh 8/26/2011 Inputs: 1)Considerations 2)Constraints 3)Weighted priorities
Page 20 Output of MCEVE Results 20 MCEVE / Abdullah Almurayh 8/26/2011 Results: 1)Considerations 2)Weighted priorities 3)List of solutions
Page 21 Details of Virtualization Solution 21 MCEVE / Abdullah Almurayh 8/26/2011 Suggestions: 1)Hardware 2)Platform 3)Measurements
Page 22 Details of MCEVE Configurations 22 MCEVE / Abdullah Almurayh 8/26/2011 Detailed configurations
Page 23 MCEVE Software Efficiency 23 MCEVE / Abdullah Almurayh 8/26/2011 Execution time: Execution time is the time between the submission and result delivering. Execution time is impacted by the quantity of the considered parameters.
Page 24 Evaluation Effectiveness: The functionality of the Model depends on the user inputs that any failure of a specific input can be effective. The Model responds to user changes and functions relatively to these changes. Accuracy: This demonstrates how precisely and accurately the Model produces the results. Compare a human perspective to the Model results Data Transparency: Data transparency in the Model indicates the data independency which exists when the code is not subjected to change when any change in the data occurs. 24 MCEVE / Abdullah Almurayh 8/26/2011
Page 25 Referenced Configuration in MCEVE 25 MCEVE / Abdullah Almurayh 8/26/2011 BenchmarksMeasurementsDescription Application Server Performance Request per second in application virtual server Web Server Performance Request per second in Web Server virtual server Mail Server Performance Request per second in Mail virtual server Overall Performance 7067 Overall Performance of the above performance benchmarks Capacity 432 Number of virtual servers per one physical server Avg Resp. Time App Server Average response time of application virtual server performance Avg Resp. Time Idel Average response time of Idle virtual server performance Price $18, The price of purchasing/licensing a virtualization platform Configuration # 14: Intel Xeon E GHz, 80 cores, 8 chips, 10 cores/chip, 2 threads/core, 2 TB RAM (128 x 16 GB, Quad Rank x4 PC CL7 ECC DDR3 1066MHz LP RDIMM), 576 x 73 GB 15k RPM SAS storage, 2 x Broadcom NetXtreme II Gigabit Ethernet, 4 x Intel x520 10Gb, RedHat RHEV Virtualization Platform.RHEV
Page 26 Impact of Weighted Priorities on Selection 26 MCEVE / Abdullah Almurayh 8/26/2011 Weighted priorities Conf#14 score Application Performance 101 Scenario (0)
Page MCEVE / Abdullah Almurayh 8/26/2011 Weighted priorities Conf#14 score Application PerformanceAvg Resp. Time App Server Scenario (1) Impact of Weighted Priorities on Selection
Page MCEVE / Abdullah Almurayh 8/26/2011 Weighted priorities Conf#14 score Application Performance Avg Resp. Time App Server Overall Performance Score Scenario (2) Impact of Weighted Priorities on Selection
Page MCEVE / Abdullah Almurayh 8/26/2011 Weighted priorities Conf#14 Score Application Performance Avg Resp. Time App Server Overall Performance Cost “price” Scenario (3) Impact of Weighted Priorities on Selection
Page MCEVE / Abdullah Almurayh 8/26/2011 Weighted priorities Conf#14 score Application Performance Avg Resp. Time App Server Overall Performance PriceCapacity Scenario (4) Impact of Weighted Priorities on Selection
Page MCEVE / Abdullah Almurayh 8/26/2011 Configuration # 14 Scores based on the users weighted priorities Lower is closer to the user’s expectation Impact of Weighted Priorities on Selection
Page MCEVE / Abdullah Almurayh 8/26/2011 benchmarks comparisons between Configurations #14 and its competitive Configurations #15 Benchmarks Conf#14 Measurements Conf#15 Measurements Application Performance Web Server Performance Mail Server Performance Overall Performance Capacity Avg Resp. Time App Server Avg Resp. Time Idel Price$18, $ 4, Conf#14 Conf#15 SPECvirt_sc2010 Result Benchmarks Comparisons of Competitive Configurations
Page 33 Lessons Learned Field Research Needed a lot of time to read and research in many different topics related to Virtualization. Needed to implement parameters prediction algorithms, but could not be validated. Tested Virtualization solutions to understand the differences between them. Testing Xenserver on UCCS HP blade servers due to unsatisfied requirements in my desktop. Needed a lot of effort for calculating converting SPEC data into data that MCEVE uses. Proposed Solution I developed a model (MCEVE ) for suggesting better solutions of virtualized environments based on the user weighted considerations. MCEVE still needs a lot of data to ensure that MCEVE yields dependable results. The execution time of MCEVE increases as the data grows The accuracy of MCEVE needs to be normalized since it is impacted by big values such as the cost. 33 MCEVE / Abdullah Almurayh 8/26/2011
Page 34 Future Work Working in future on a different evaluation method by testing the suggested configurations. Identifying fundamentally different opportunities to provide vast data comes from trustworthy sources The Model needs a Data Standard that includes naming agreements for data elements and other system components. There is a need for an offset that can be associated to a large benchmark such as the “Price” to reduce its negative effectiveness. The Model application can be published in the real world and surveyed. 34 MCEVE / Abdullah Almurayh 8/26/2011
Page 35 Conclusion I proposed the MCEVE Model that could help users to efficiently configure virtualized environments in an easier and reduced cost manner. The model uses user considerations and configurations as inputs for algorithmic outputs /suggestions. The proposed Model helps to minimize unexpected events driven by inefficient configurations. The model saves user’s time by adopting the right decisions in hours instead of spending days researching and comparing existing results and reduces the cost of performing tests and experiments. Performance can be enhanced by using solutions accordingly to real solutions rather than trying different solutions that can lead to high cost. 35 MCEVE / Abdullah Almurayh 8/26/2011
Page MCEVE / Abdullah Almurayh 8/26/2011 Demo and Questions
Page 37 Bibliography 37 MCEVE / Abdullah Almurayh 8/26/2011 [1]VMware, "Virtualization Overview," [Online]. Available: [2]VMware, "Disaster Recovery Virtualization," [Online]. Available: [3] Wikipedia, "Comparison of platform virtualmachines," 7 July [Online]. Available: [4] Intel, "Intel® Virtualization Technology (Intel® VT)," [Online]. Available: [Accessed 1 Jan 2011]. [5] AMD, "AMD Virtualization (AMD-V™) Technology," [Online]. Available: [Accessed 11 Jan 2011]. [6] Wikipedia, "Comparison of application virtual machines," 15 May [Online]. Available: [7] J. P. Walters, Vipin Chaudhary, Minsuk Cha, Salvatore Guercio Jr. and Steve Gallo, "A Comparison of Virtualization Technologies for HPC," in 22nd International Conference on Advanced Information Networking and Applications/DOI /AINA , [8]VMware, "A Performance Comparison of Hypervisors," [Online]. Available: [9] P. Barham, B. Dragovic, K. Fraser, S. Hand, T. Harris, A. Ho, et a, "Xen and the art of virtualization," in In Proc. Of the 19th ACM Symposium on Operating System Principles, Bolton Landing, NY, Oct [10] S. Nanda and T.-c. Chiueh, "A Survey on Virtualization Technologies," [Online]. Available: [Accessed 05 Dec 2010]. [11] IBM, "Virtualization — why it's hot and how to get started," [Online]. Available: [Accessed 09 Dec 2010]. [12]White, J., & Pilbeam, A., "A Survey of Virtualization Technologies With Performance Testing," [Online]. Available: [13] A. Metelo, "Not All Server Virtualization Solutions Are Created Equal," IBM SWG Competitive Project Office, 13 Aug [Online]. Available: ftp://ftp.software.ibm.com/software/solutions/2982/Not_All_Server_Virtualization_Solutions_Are_Created_Equal.pdf. [14]J. Fornaeus, "Device hypervisors," in Design Automation Conference (DAC), th, Alameda, CA, 18 June [15]C. Scheffy, Virtualization For Dummies,® AMD Special Edition, AMD Special Edition ed., Hoboken, NJ: Wiley Publishing, Inc., 2007, pp [16] Binbin Zhang, Xiaolin Wang, Rongfeng Lai,Liang Yang, Yingwei Luo, Zhenlin Wang and Xiaoming Li, "A Survey on I/O Virtualization and Optimization," in The Fifth Annual ChinaGrid Conference/DOI /ChinaGrid , [17]I. Habib, "Virtualization with KVM," 01 Feb [Online]. Available: [18]Redhat, " [Online]. Available: [19] M. T. Jones, "An overview of virtualization methods, architectures, and implementations," 29 Dec [Online]. Available: [20] J. Fisher-Ogden, "Abstract Hardware Support for Efficient Virtualization," 12 Dec [Online]. Available: [21]R. Arash, H. Salimi and M. Sharifi, "Improving Software Dependability Using System-Level Virtualization: A Survey," in 2010 IEEE 24th International Conference on Advanced Information Networking and Applications Workshops, IEEE, 2010.
Page 38 Bibliography 38 MCEVE / Abdullah Almurayh 8/26/2011 [22]B. Underdahl, M. Lewis and T. and Mueting, Cloud Computing Clusters For Dummies, AMD Special Edition ed., Hoboken, NJ: Wiley Publishing, Inc., [23] I. G. Education, "Virtualization in Education," October [Online]. Available: 07.ibm.com/solutions/in/education/download/Virtualization%20in%20Education.pdf. [24]P. X. Zhou, "Distributed and Internet Systems Lab," [Online]. Available: [Accessed 07 May 2011]. [25]C. E. Chow, "Homework #3. LVS Cluster," 09 March [Online]. Available: [26]T. L. V. Server, "What is virtual server?," 13 Jan [Online]. Available: [27]Amazon, "Amazon Elastic Compute Cloud," [Online]. Available: [Accessed 23 May 2011]. [28] PlateSpin, "Consolidated Disaster Recovery Using Virtualization," [Online]. Available: [29]Xen, "What is Xen Hypervisor?," 03 March [Online]. Available: [30]V. Inc., "Virtualization Basics," [Online]. Available: [Accessed 19 May 2011]. [31]M. Tulloch, Understanding Microsoft Virtualization Solutions, 2nd ed., Microsoft Press, 2010, pp [32] M. Ahmed, S. Zahda and M. and Abbasbas, "Server Consolidation Using OpenVZ: Performance Evaluation," in II th International Conference on Computer and Information Technology, December, [33]K. Kolyshkin, "Virtualization in Linux," 1 September [Online]. Available: [34] R. HAT, "RED HAT Enterprise Virtualization Hypervisor," 29 October [Online]. Available: [35]OpenVZWiki, "Virtuozzo," 18 June [Online]. Available: [36] P. V. Containers, "An Introduction to OS Virtualization and Parallels Virtuozzo Containers," 27 April [Online]. Available: [37]VirtualBox, "Welcome to VirtualBox.org!," Oracle, [Online]. Available: [Accessed 22 March 2011]. [38]Oracle, "Oracle VM VirtualBox," Oracle, [Online]. Available: [Accessed 22 March 2011]. [39] Linux-VServer.org, "Welcome to Linux-VServer.org," Linux-VServer.org, 17 March [Online]. Available: VServer.org. [40] D. Gelernter, "Truth, Beauty, and the Virtual Machine," DISCOVER, 1 September [Online]. Available: [41]Wikipedia, "Virtual Machine," 12 July [Online]. Available: [42] SUSE, "SUSE Linux Enterprise Server Virtualization with Xen," Novell, Inc, 03 Dec [Online]. Available: [43]Wikipedia, "Virtual private server," 15 July [Online]. Available: [44]VMware, "Virtualization Basics," VMware, Inc., [Online]. Available: [45]Casteleyn, Sven, Florian Daniel, Peter Dolog, and Maristella Matera, Engineering Web Applications, Springer, 2009, p The WebML Model..
Page 39 Bibliography 39 MCEVE / Abdullah Almurayh 8/26/2011 [46]SPEC, "SPECvirt_sc2010 Results," [Online]. Available: [47] SPEC, "Performance Details," [Online]. Available: perf.html#Performance Details. [48] VMware, "VMware Cost-Per- Application Calculator Methodology," [Online]. Available: calculator- methodology.pdf. [49] VMware, "VMware vSphere™ 4.1 Pricing, Packaging and Licensing Overview," August [Online]. Available: [50]P. Rob and C. Coronel, Database Systems: Design, Implementation, and Management, 8th ed., Cengage Learning, 2007, p. 704 pages. [51]Apache, "Apache Web Server," [Online]. Available: [52] Michelle J. Gosselin, Jennifer Schommer, "Confining the Apache Web Server with Security-Enhanced Linux," [Online]. Available: [53]W3Schools, "HTML Tutorial," [Online]. Available: [Accessed 20 June 2011]. [54]Wikipedia, "PHP," [Online]. Available: [55]P. D. Group, "PHP Manual," July [Online]. Available: [56]Wikipedia, "MySQL," 12 July [Online]. Available: [57]MySQL, "MySQL 5.5 Reference Manual," Oracle, [Online]. Available: [58]Wikipedia, "Solid state drive," [Online]. Available: [59] Ekker, Neal; Coughlin, Tom; Handy, Jim, "The Solid State Storage," January [Online]. Available: [60] IBM, "SPECvirt_sc2010 Result," SPEC, Apr [Online]. Available: perf.html. [Accessed May 2011]. [61] IBM, "Configurations #15 - SPECvirt_sc2010 Result," SPEC, Jul [Online]. Available: perf.html. [Accessed Aug 2011]. [62]Weibull, Experiment Design and Analysis Reference, ReliaSoft, 2008, p [63]&. A. P. John L. Hennessy, Computer Architecture, Morgan Kaufmann ed., [64]XEN, "The Xen® hypervisor," [Online]. Available: [65]Citrix, "XenServer," Citrix Systems, Inc, [Online]. Available: [66]Citrix, "XenCenter," Citrix Systems, Inc, 07 March [Online]. Available: [67]R. Cordova, "Enhancing Network Scanning For Discovering Vulnerabilities," University of Colorado at Colorado Springs, Colorado Springs, CO, [68]T. j. Project, "jQuery," The jQuery Project, [Online]. Available: [Accessed 2011 Aug 03].