Comparison of Technologies for Software ization PETR SUBA, JOSEF HORALEK, MARTIN HATAS Faculty of Informatics and Management, University of Hradec Králové, Rokitanského 62, 500 03 Hradec Kralove Czech Republic josef.horalek@uhk.cz, petr.suba@uhk.cz, martin.hatas@uhk.cz http://www.uhk.cz Abstract: The article presents the results of testing of some selected freely available tools for software virtualization. The article introduces at first the basic terms and principles of virtualization and the potential of its use. It further deals in more details with the full virtualization principles focused to the software virtualization. The article then presents the selected tested tools and testing methods. The measured values are represented by means of graph and its interpretation. On the basis of the measured data, one of the tested tools with the best proved properties of the measured values is recommended. Key-Words: ization, Windows PC, Oracle Vistual Box, CPU, memory, test, usability 1 Introduction to virtualization The aim of this article is to present the basic principles of virtualization and to recommend a tool for software virtualization on the basis of tests performed with some selected freely available virtualization tools. The given tool will be then used at the Faculty of Informatics and Management, University of Hradec Králové (FIM UHK) for specific operating system tasks testing. ization as a concept was developed for the first time by IBM in 1960s. During 1980s and 1990s, desktop computers and x86 servers became much more available and, therefore, virtualization had not undergone any significant progress at that period. The client-server applications and emergence of operating systems Windows and Linux made both the server technologies and home PCs markedly cheaper. However, new challenges have appeared, such as time-consuming maintenance of the existing computers, high operating cost or high failure rate, which led to the development of x86 virtualization. ization has increased the effectiveness dramatically and reduced the overall IT expenses. At present, virtualization is understood to be a relatively new concept by means of which many practical problems related to the information technologies use can be solved [1]. In general, we can say virtualization is a means allowing creation of the so-called virtual objects. ization can be performed on various levels, from the whole computer to the individual components creation: We can have a virtual disc, a virtual memory, or a virtual processor. By means of a virtualization tool, a different operating system can be started in an ing operating system. ization changes nearly all aspects of the operating system and storages administration. Using the virtualization technology, we can create training, testing, or even developmental environments; there are more possibilities to save physical means and expenses. And we must not forget another important advantage of the virtualization technology resistance to failures and high availability. ization allows more operating systems run in one physical machine, e.g. Windows, Linux etc. The number of virtual machines that can run simultaneously in one machine depends on hardware specification. At present, virtual machines are implemented in most data centres as normal servers but with much less maintenance and administration cost. This technology potential is huge and plays a great role in the computer science future [2]. 1.1 ization benefits Consolidation of servers definitely is the greatest benefit of virtualization. ization has become a saving means of about 60% of companies. Their IT departments have been transforming the physical servers to the virtual ones. Such transformation has several advantages [3]. Continuity of operation and high availability The virtual server is separated from a particular physical hardware and can easily be backed up and restored in another hardware or another locality. Advanced virtualization technologies thus allow the transfer of a virtual server to another physical server without disconnection of working users, even in applications not supporting the high availability solution at all. Scalability The separation of the physical hardware from the individual virtual servers and stations allows an easy increase of the applications performance by a simple ISBN: 978-1-61804-028-2 160
addition of a new hardware in the existing infrastructure. If, for example, the number of demands made on a server is increased, the memory or CPU performance can simply be increased in the administration console. The same it is with the addition of any other hardware allowed by the given virtualization technology. Attendance simplification and cost of operation reduction Thanks to the reduction of the number of physical servers, automated installation of new servers from templates, their easy transferability to another hardware, and thanks to the centralized administration tools, the attendance activities during the servers and applications installation and operation have been simplified significantly. More effective administration The symbiosis of a company IT environment and its commercial needs should be the primary aim of the consolidation. The aim of the IT consolidation is mainly to shorten the period of the return on investment in the IT infrastructure and to reduce the expenses related to IT administration. The administration of an increasing number of servers, working stations and other devices of the company IT infrastructure can be expensive and demanding from the capacity point of view. Working stations consolidation eliminates the autonomous administration disadvantages to a great extent. IT departments can administer and maintain the software on such stations much more effectively, including the central application of safety patches and correction packets. A unique, consolidated and consistent environment in a company can increase the employees labour productivity and reduce the expenses. A unique setting and compliance with the configuration policy guarantees much less vulnerability of the centrally administered stations towards the safety threats [2, 3]. 1.2 Full virtualization The terminology and classification of the individual virtualization types is not uniform. For the purposes of this article, the classification published in [3] was used. ization can be classified according to the principle used to a full virtualization, simulation and emulation, paravirtualization, virtualization of storages, virtualization of networks, and virtualization of applications. From the point of view of the tested tools, the full virtualization principles were used. Under the given conditions, it is the most common type of virtualization when all parts of the computer are virtualized. The operating system running in this environment does not recognize that it has no access to the physical components of the given machine. The physical layer is divided from the programme layer. All programmes are thus running on a virtual hardware. The advantage of such virtualization is that the virtual components parameters can be defined (capacity of the disc, memory, etc.) and the programmes are independent on the given technical equipment. No hardware change influences the virtual environment as shown in Figure1. Another advantage is that the physical hardware need not be connected to the virtual environment. This means that in full virtualization, the programmes or operating system can be started in other processor platforms as well, which results in full transferability. In addition to the above mentioned advantages, this type of virtualization has also some disadvantages. Paradoxically, it is just the division of the physical and programme level which is the greatest disadvantage because in such case, full performance of the virtual machine cannot be reached although the virtual hardware is an image of the physical one. The reason is that various activities like operations with the memory, processor instruction, or access to the disc are emulated and not performed directly by the physical hardware. Fig.1 Principles of full virtualization Full virtualization allows the use of two architectures, the ed architecture or the hypervisor one. The ed architecture uses the already running operating system in which virtualization-allowing software has been installed, for example Windows PC or Box. The other variant is to install the hypervisor directly in the hardware. It then becomes the virtualization layer running under the virtual machines, for example the VMware ESX server. The operating system ed architecture is called the software ISBN: 978-1-61804-028-2 161
virtualization while the hypervisor architecture is called the hardware virtualization. 2 ization tools selection For the purposes of the specified area of interest, we had to select such tools that are easy available, their machine equipment requirements are feasible, and can be operated even without any advanced knowledge of virtualization technologies. As to the level of equipment by computer technology, the environment of a common university in the Czech Republic is very similar to that in both the production practice and state administration. The assessed average age of a user station is from two to four years. Typically, the MS Windows XP or 7 is installed on it. Such system is administered by a separate entity who is interested in a stable and nointervention operation. In general, such environment is not suitable for experimental purposes and virtualization provides a compromise solution between the time, safety and lifelikeness of experiments. A set available commonly in the informatics laboratories and classrooms has been chosen for the selection of usable technologies. The limitation given by the use of a common work station limited, to some extent, the selection of offered technologies to those with lesser equipment demands and, therefore, we had to exclude the technologies requiring the inaccessible server technologies. For these reasons, two freely available representatives of the software virtualization were chosen for the testing, namely Oracle Box and Microsoft PC. Both selected technologies were used in default condition without any intervention in the configuration. 2.1 Tested values selection To allow an objective comparison, we had to choose such approaches that would represent sufficiently the burden during the use in practice and, at the same time, would have sufficient informative value as to the performance and direction of the individual technologies. We chose the values representing the computing performance of the processor, speed of reading and recording on the virtualized hard disc, and the network interface from the point of view of its attendance and throughput. All tests were performed with maximum consideration of constant conditions of the running test. Therefore, the same was always used and the ed system was restarted after each measurement. The tested procedure attendance was always provided by a debugged script. If it was possible, the given task was tested in the system as well. The system results were used as reference values. The processor (hereinafter CPU) was tested in a way allowing the test performance in all tested technologies and comparison and objective assessment of results. For this reason, we chose the PI number computing to a constant number of decimal digits and encrypting of a testing set of random symbols by the OpenSSL library. The speed of the access to the disc, i.e. to its virtual image, was tested in form of the duration of a 1 GB file copying from one location to another within the same disc image. Thus both the recording and reading were tested concurrently. In the network interface, we tested the measurement of the ed system response from the external source of requests (echo request, echo reply). 2.1.1 Tests composition and technical parameters CPU encryption test Copying of a randomly generated 1 GB file was tested. The sequence of commands in the utility script was as follows: (1) system boot up and user login; (2) 90-second waiting till the system procedures stabilization; (3) start of the file copying from the root directory to another place as the Timeit.exe (Time for Linux) parameter; (4) Timeit.exe (Time for Linux) output recording in a text file after the copying termination; (5) the copied file deletion; (6) system restart. The whole procedure was repeated 100 times. CPU test π computing Π computing to 8,388,608 (8M) decimal points was tested. The SystemStabilityTester programme was used for the computing, using two non-cooperating links and Gauss Legendre algorithm. The sequence of commands was as follows: (1) system boot up and user login; (2) 30-second waiting till the system procedures stabilization (3) π computing; (4) the computing output recording in a text file after the copying termination; (5) system restart. The test was repeated 100 times and the results were recorded in two forms, directly from the calculating programme and from Timeit.exe (Time for Linux). HDD test Copying of a randomly generated 1 GB file was tested. The sequence of commands was as follows: (1) system boot up; (2) 90-second waiting till the system procedures stabilization; (3) start of the file copying after logging-on as a parameter of Timeit.exe (Time for Linux); (4) Timeit.exe (Time for Linux) output recording in a text file after the copying termination; (5) the copied file deletion; (6) system restart. The test was repeated 100 times and Timeit.exe (Time for Linux) was used both for values recording and measurement. ISBN: 978-1-61804-028-2 162
NIC test After the network interface was configured and the network put into operation, we tested the duration of response on the network layer of the ed system. Ping, a generally known programme, was used for testing. The requests were sent from an external source and were recorded in a file. The request was sent 100 times and the result was recorded. 3 Measured data As we mentioned above, Windows 7 32b is the primary operating system at FIM UHK. Therefore, this system was chosen to be the operating system (OS). The OS was tested by means of the same series of tests as the virtualized machines and its values were used as the reference ones. It is called below in the text. Windows 7 32b and Squeeze 6.0.0. were the ed systems. Box a PC, the pair mentioned above already, were the virtualization environment. Four tested environments thus originated, called and on for systems virtualized on the PC, and on and on for systems virtualized on the Box. NIC test In this test, we measured the time response from the local server to the virtualized NIC by means of the Ping programme. The server and the were connected directly by means of a gigabyte crossover UTP 5E cable. The results are shown in seconds. on on Average 0.8398 0.8265 0.4903 0.4179 0.1884 deviation 0.055 0.168 0.107 0.055 0.013 Variance 0.003 0.028 0.011 0.003 0.000 Maximum 1.08 1.98 1.09 0.79 0.231 Quartile 0.856 0.823 0.516 0.437 0.193 Median 0.844 0.793 0.479 0.4185 0.185 1st Quartile 0.835 0.774 0.448 0.395 0.181 Minimum 0.429 0.393 0.285 0.319 0.171 Table 1 Results of the NIC test It is evident from the measured values that the least time delays were shown during the NIC test by the virtualized system by means of the Oracle Box. Box showed better results during the virtualization by means of the OS as well. CPU test π computing The test was aimed to the computing performance of the virtualized operating system. Data acquired by the SystemStabilityTester programme in all tested environments were used. The values are shown in seconds. on on Average 336.2818 489.9211 447.6452 493.5915 183.4244 deviation 24.986 14.595 1.709 3.929 0.474 Variance 624.323 213.023 2.920 15.436 0.225 Maximum 428.495 497.438 450.718 503.499 185.219 Quartile 329.227 494.560 448.783 495.540 183.643 Median 324.513 491.599 447.950 492.947 183.308 1st Quartile 324.165 489.336 447.083 490.820 183.140 Minimum 316.48 369.32 443.97 483.64 182.46 Table 2 Results of the CPU test π computing It follows from the measured values that Windows PC showed better results in virtualization of both operating systems in the tests focused to CPU computing performance. CPU encryption test In the encryption-focused tests, the time required for a randomly generated 1 GB file encryption was measured. A 256b AES encrypting algorithm was used for the encrypting. The values are shown in seconds. The encryption results were not stored to avoid the influence of the data recording speed. on on Average 18.511 21.362 20.982 26.523 12.428 deviation 2.167 0.809 0.790 1.374 1.085 Variance 4.697 0.654 0.624 1.887 1.177 Maximum 28.390 25.330 22.221 32.506 14.164 Quartile 18.233 21.408 21.300 26.846 13.498 Median 17.955 21.250 21.000 26.227 11.559 1st Quartile 17.445 21.100 20.289 26.015 11.540 Minimum 17.17 18.78 17.62 18.10 11.51 Table 3 Results of the CPU encryption test These tests also show that in CPU use, virtualization by means of Windows PC is more effective. ISBN: 978-1-61804-028-2 163
HDD test In this test, we measured the duration of the data recording and reading on a virtualized 1 GB HDD. The values are shown in seconds. requirements were stressed, the Microsoft solution should be preferred. On the other side, the Oracle product can be recommended in all situations where migration to another platform can be expected. on on Average 53.806 40.613 32.028 32.588 27.933 deviation 2.561 1.114 0.688 1.174 1.554 Variance 6.560 1.240 0.474 1.377 2.415 Maximum 62.890 42.240 33.408 37.231 32.510 Quartile 54.980 41.313 32.797 33.057 28.618 Median 53.580 40.655 31.915 32.604 27.979 1st Quartile 51.990 40.235 31.443 31.625 27.338 Minimum 47.99 35.64 30.94 30.94 24.91 Table 4 Results of the HDD test Marked differences were measured during the recording and reading testing in OS virtualization on PC and Oracle Box. The measured results suggest that Oracle Box access to the virtualized HDD is markedly quicker. In the Windows 7 operating system virtualization, the results do not differ markedly although, when comparing the measured values, Windows PC is quicker in a nonsignificant extent. 4 Evaluation Several conclusions can be made from the results of our measurements. The results are very favourable for the needs of the environment from which the request for such research came. We can say that despite evident differences between the individual tools, both are usable very well and meet the educational requirements. Both virtualization tools can be used for training of the tasks related to the administration and protection of both the operating systems and computer networks. In both cases, the differences between the native and virtual machine reached no more than the multiples of the reference values measured on the system. These facts are shown in Figure 2. It shows a comparison of the differences of values related to the reference values; the system result is considered to be 100%. Therefore, both virtualization technologies can be recommended for the educational needs. If the performance Windows7 PC Box Linux PC Box Windows 7 HDD 32.0277 32.5879 53.8060 40.6126 27.93299 CPU Encryptin g 20.9822 26.5234 18.5105 21.3617 12.42844 CPU Computin g PI 447.645 2 493.591 5 336.281 8 489.921 1 183.42439 NIC Echo 0.4903 0.4179 0.8398 0.8265 0.18837 References: [1] DEDLEY, H. History of ization. Infobarrel.com [online]. 2009, 1, [cit. 2011-06- 07]. Available from WWW: <http://www.infobarrel.com/history_of_virtua lization>. [2] MARSHALL, D., The Rise of Systems and ization (on-line), izationsdefrag.com Editors, 2009 [cit. 2011-06-07]. Available from: <http://www.virtualizationdefrag.com/articles/h istory-of-virtualization/index.php>. [3] RUEST, N., RUEST, D., ization - The beginner s guide, Mc. Graw Hill, New York, 2009, Edit. 1, 442 pp., ISBN: 978-07-161401-6. [4] TULLOCH, M. and MICROSOFT VIRTUALIZATION TEAMS Understanding Microsoft ization Solutions: From the Desktop to the Datacenter. 2. Redmond : Microsoft Press, 2010. 466 pp. Library of Congress Control Number: 2010920178. ISBN: 978-1-61804-028-2 164
Fig.2 Final comparison ISBN: 978-1-61804-028-2 165