In a previous work [ 4
], we tested the possibility of using Virtualization
(Xen) [ 5
] and Virtual Network Computing
(VNC) [ 6
] technologies to build a remotely accessible laboratory system that is suitable for Computer Networks
courses introducing concepts related to network addressing, routing, traffic flow-controlling, etc. Virtualization
technology was used to create a network comprising a number of Virtual Machines
(VM) on a single server as shown in Fig. 1
, and in these VMs run real Operating System (OS) instances. With the capabilities modern OSs have, these instances can perform the functions of all network devices such as Routers, IDSs, IPSs, etc. The machines have been set up to enable remote access and learners were able to connect to their consoles through the VNC modules installed on the host machine and on the VMs in order to complete some practical exercises.
Fig. 1. A network constructed of a number of VMs on one machine.
Manually creating the VMs and setting them up for remote access involved lot of effort and was a time-consuming process that instructors were requested to carry out whenever the network was to be rearranged to adopt modifications necessary for introducing new concepts. Furthermore, learners were not able to modify the network structure in order to try scenarios different from the ones the instructors have created. And because certain settings were required for enabling remote access to the VMs, these settings were prohibited from being modified. All these issues brought forth the idea of having a tool for automating the creation of the network, and finding a way to allow learners work on the devices with no restrictions.
1.1 Research Contribution
In this paper, we demonstrate how we used Xen and VNC technologies to build a modular Web-based laboratory (NVLab) suitable for ICT courses that require the individual learner to work on multiple networked devices. Learners will use the tools NVLab includes to produce a network design and then construct it on the server. After this phase is completed, learners will be able to connect to the server and work on their networks. In our laboratory, the VMs need not be preconfigured for remote access, and learners are able to work freely and with full control over them. NVLab also enables the instructors to produce properly/improperly configured network designs and ask the learners to verify/troubleshoot the settings of the devices included in these networks.
The rest of the paper is organized in sections as follows: Section 2 provides brief background information, and Section 3 presents description of related work in the field of ICT Web-based courses. Then, in Section 4, we introduce the core technologies used in NVLab. A description of how the system was implemented is taken up in Section 5. Next, we introduce a case study in Section 6. Finally, we conclude with our results in Section 7.
As in all engineering curricula, laboratory activities play a major role in ICT courses. These activities not only help learners get better understanding of what they are being taught, but also provide them with the problem-solving skills and experience necessary for adopting their knowledge in real-life contexts [ 1
], [ 2
]. Beyond this, laboratory activities became the main components of the course rather than a separate supporting part, and it is now common to find laboratory-based courses delivered at increasing number of universities and training centers [ 3
Among ICT courses that require the learner to work on multiple devices is the Computer Networks course. These days, computer networks have widened to cover the whole world facilitating communication between individuals, access to various resources, e-commerce, etc. That is a reason why well-trained and qualified staff personnel are required at increasing number of companies not only specialized in installation and maintenance, but also in other fields. It is now common to find Network Experts working at stores, government offices, transport stations, and at every place where information is to be exchanged. Consequently, the need for training centers/laboratories utilizing proper infrastructure necessary for qualifying individuals to be Network Experts arises.
Configuring devices operating at different levels of the OSI standard, maintaining connectivity between these devices on one hand and with the external world on the other hand, traffic and load balancing, and upholding carefully designed security policy are among the network administration tasks. Hence, a laboratory addressing Computer Networks issues should contain enough machines of various types, computers, switches, routers, firewalls, IDSs, IPSs, etc., to cover all the above mentioned tasks, among others. Besides, every individual learner should have access to a set of these devices in order to try different possible network scenarios that meet the requirements and pick one of them to be implemented.
While on-campus facilities are designed to meet the previously mentioned criterion, Web-based laboratories are evolving to become equal. This has been reinforced by the introduction of emerging technologies and the increased demand for them to be used within the e-learning environments. Systems implementing emerging Virtualization
and Remote-Access technologies to deliver Computer Networks
training have been successfully tested and verified to be adequate for reaching the goal [ 23
], [ 24
Work dealing with building a Web-based laboratory environment is of wide range and is not limited to courses related to ICT. But, the design of the laboratory differs according to the field of application and the availability of resources. In this regard, a number of methods have been investigated.
The first approach suggests equipping the on-campus laboratory facilities with remote access network. In this way, learners can access and work on real devices while monitoring the results of the modifications they do to the system. This arrangement has been adopted in courses of wide range such as: electronics engineering [ 14
], robotics [ 15
], besides ICT courses [ 16
], [ 17
]. A drawback of this method is the requirement to have a suitable preinstalled laboratory facility and more resources for the remote access devices. Moreover, changing the configuration of the laboratory requires the presence of some staff on the laboratory's premises or complex switching arrays [ 18
Other researches proposed to use Web-based software simulation tools within the learning environment. Software simulators are rich with animations and variable controls that help learners produce a more meaningful mental model for the introduced concepts. A combination of simulators designed in this way and static materials (text, images, etc.) have proved to be very effective in reaching a better learning performance with medical science students [ 25
]. But, according to a recent research, special care should be given to the complexity of the contents these tools introduce [ 26
]. In other cases, these tools are designed to imitate a hardware device and provide an easy and flexible way of exploring it. In engineering, Web-based laboratories using tools designed in this way are of wide range, [ 19
] in electronics engineering, and [ 20
], [ 21
] in ICT-related courses. In the software simulators, the laboratory structure can be easily altered and the cost is significantly reduced especially when implementing Open Source code programs. However, a disadvantage of such systems is that work is done on a simulated device which puts limitation on what can be tested since simulators are not designed to cover all aspects of the real device.
The introduction of new technologies for implementing and remotely accessing various devices evoked a new way in designing Web-based laboratories especially in the field of Computer Science
technology, for example, is now being used to create several OS instances and run them on one host machine [ 22
]. In parallel, Remote Desktop
, Secure Shell
, Virtual Network Computing
technologies, etc., serve in facilitating access to the laboratory through the Web. Systems implementing these technologies to deliver Computer Networks
training have been successfully tested and verified to be adequate to for reaching the goal [ 23
], [ 24
]. They offer the learners real instances of the laboratory devices to practice on, and feature flexibility and mobility. However, in most cases, these laboratories served to provide training on a narrow range of computer networks concepts such as Network Security
, for example.
Introduction of software virtualization technology goes back to the sixties of the last century. Since that time, it passed through periods when its importance faded out; however, new fields of application have emerged and the technology is strongly back to the scene [ 7
Inserted as a software layer on top of the host machine's hardware, the Virtual Machine Monitor (VMM) allows transparently sharing the resources among multiple guest OS instances which run while isolated from each other, but they still can communicate via the different software networking scenarios the VMM offers.
Software vendors introduced a number of modes in which the VMM builds and controls the VM environment [ 8
]. For example, VMs can be run in the full-virtualization
mode in which they have no access to the host machine's resources; instead, the VMM simulates the functions of all these resources. On the other hand, VMs running in the paravirtualization
mode can access some hardware parts in the host machines. Differences between these techniques come in terms of overhead imposed by inserting the new layer between the hardware and the guest OSs, and the simplicity of implementation. In this regard, the full-virtualization
mode is easier to implement but has a less performance indicator; while the paravirtualization
mode has a better performance figure but requires inserting special modifications in the guest OS, i.e., it is difficult to implement and with no support in legacy OSs.
Open Source Xen [ 9
] is one software VMM (called Virtualization Hypervisor
in Xen terminology) offering full-virtualization
techniques. Its free version has been included along with a set of supporting tools in a number of OS distributions, CentOS, for example. In Xen terminology, the guest-OSs are called Domains
with a special privilege given to Domain-0
in which runs the hypervisor. This privileged domain will handle the resources provided to the running VMs.
Xen introduces a number of networking scenarios in which the VMs can exchange traffic among each other, on one hand, and with the external world (including Domain-0)
, on the other [ 10
]. These scenarios include: bridging
, and Virtual Network
. In all these scenarios, a Virtual Interface
will be bounded to each NIC the VM has (including Domain-0
) and will be used to pass the traffic to/from this network interface. The pair consisting of the VM's interface and the Virtual Interface
bounded to it can be thought of as two Ethernet interfaces connected by an internal crossover cable. Domain-0
will use standard OS mechanisms for bridging, routing, rate limiting, etc., to handle the traffic to/from this Virtual Interface
When a bridged network mode is selected, a VM can be connected internally to any software switch, or more than one software switch that Xen creates. The software switches work in the Network layer (L2) of the OSI model directing traffic based on the MAC address assigned by Xen. Routed network, in contrast, creates routes between Domain-0 and the other Domain(s) in the system allowing traffic to be routed based on the IP address. Another possible networking configuration in Xen is the Virtual Network, but it is not standard.
Providing an easy to use graphical user interface ( Fig. 2
), Virtual Machine Manager
]) is a free open-source support tool for managing the VMs in Xen.
Fig. 2. Virtual Machine Manager.
Using this tool, tasks like creating new VMs, modifying the hardware resources attached to them, booting/shutting down VMs, and monitoring their performance and resource utilization can be carried out.
creates for every VM installed on the host machine a configuration file that contains entries of hardware resources assigned to it such as the number of virtual CPUs, the size of memory, storage type and size, network interfaces, etc. By modifying the contents of this file using any simple text editor, it is possible to alter some of the resources assigned to the VM to which this file belongs. Sample configuration file of a VM named "R" is provided in Fig. 3
Fig. 3. Virtual Machine's configuration file.
In this file, the resources assigned to the VM are:
One virtual CPU ( vcpus = 1).
64 MB of memory ( memory = 64).
A regular file on the hard disk to be used as a storage medium
Two network interfaces:
(vif = ["mac=00:16:3e:78:d9:01,
bridge=xenbr1,type = ioemu",
"mac = 00:16:3e:78:d9:02,
bridge = xenbr2, type=ioemu"]).
Other information required for the proper handling of the VM's console and booting options are also included in this file.
provides a graphical console window ( Fig. 4
) as a replacement for the VM's monitor. This console can be brought to view by double clicking on the desired VM's name in virt-manager
window, then pushing the "Run" button from the console's window. The user will use the console window and the host machine's input devices to boot, pause, resume, shutdown, and interact with the guest OS.
Fig. 4. The Virtual Machine's console window in virt-manager.
4.2 Virtual Network Computing
VNC technology, originally introduced by "Olivetti and Oracle Research Laboratory, Cambridge Univ." enables users to access centralized resources remotely from widely available devices [ 12
]. Its architecture is based on a server-client model ( Fig. 5
) where the server is the VNC-Server that resides on the machine hosting the resources to be accessed, and the VNC-Client (called vncviewer) is an ultrathin or generic software.
Fig. 5. VNC architecture.
Underlying VNC is the Remote Frame Buffer
(RFB) protocol: a simple protocol for remotely accessing a GUI [ 13
]. Using frame buffer resulted in the applicability of this protocol in all OSs implementing windowing such as Windows, Linux, and MAC OS. This protocol defines subprotocols for exchanging the data between the server and the client sides, which are: the Display
, the Update
, and the Input
subprotocols. The Display
portion of the protocol is used to update a rectangle area located at (x, y) point in the client machine's frame buffer. The Update
protocol is client-event-driven, i.e., update data are sent from the server in response to a request from the user. On the other hand, within the Input
portion of the protocol, the viewer sends input events to the server whenever a keyboard or pointing device event occurs on the client's side while the viewer window is active. The VNC-server will create a display identified by a unique number for each user allowed to connect via VNC and the connection will, by default, be over TCP ports 5800+Display# and 5900+Display # (Display# is interpreted as the display number through which the remote user will login).
It is possible with VNC to simultaneously connect a number of clients to the VNC-server while keeping them stateless, i.e., closing the connection can be done from either side without undesired consequences; still, upon reconnection, the user will be able to complete the session from where it was closed. Furthermore, VNC does not require running the server and the client applications on the same Operating System, i.e., users accessing the server need not be running the same Operating System installed on the server on their machines.
With the introduction of Java as a major programming language for Internet applications, VNC-server was implemented to accept HTTP connections, and VNC-client software (vncviewer) was implemented as a Java Applet
( Fig. 6
). In this way, any Java-capable Web browser can be used as a VNC-viewer.
Fig. 6. VNC-viewer applet.
The need to supply each learner with a Web-based space offering a set of devices to try and test different network scenarios was the main reason behind introducing NVLab. Further than this, it is required to have the proper tools that enable learners construct and access the networks they want to test. In the next part, we describe how Xen and VNC were implemented, and the different tools provided to the learners in the suggested system.
5.1 Host Machine Specifications
During test phase, NVLab has been hosted on a rack-mounted server with Pentium D CPU running at 3.0 GHz, 224 GB of SCSI storage, and 8 GB RAM module. The Operating System running on the server was CentOS5 and the Web hosting application was apache tomcat 5.5.25.
For ease of installation and less need for high performance, we set up prototype VMs for a Router and PC (the currently supported device-types) machines to run in the full-virtualization
mode. From these prototypes, all machines in the network diagram will be copied. A VM is assigned one virtual CPU, 64 MB of RAM, and a number of network interfaces depending on its type (one interface for the PC machine and up to three interfaces for the Router machine). This information will be saved in the configuration file of each VM created in the system. As for inter-networking, a bridged network
configuration has been adopted in NVLab for simplicity. We created a number of software switches in the system using a script similar to the one provided in Fig. 7
and made a number of them available for each learner to use. The user will not have the privilege to modify the settings of these switches, but will be able to connect other devices to them. This will not have any impact on the exercises since Switches can normally be operated with their factory settings. For security reasons, no external traffic was allowed in the network the learners create.
Fig. 7. The script used to create four software switches (xenbr ) in Xen.
The VNC-server is configured to accept incoming HTTP connections initiated from Web browsers over VNC default TCP ports and provide the learner with an
-size applet showing the host machine's desktop upon successful verification of the login information as shown in Fig. 6
. During test stage, one user account has been created on the host machines for users connecting via VNC-server.
NVLab is realized as a Web-based environment offering tools for: producing a small-scale computer network design, creating the network by instantiating one VM for every router and PC used, and configuring the different VMs by connecting to the machine hosting them. A diagram of the system structure is shown in Fig. 8
Fig. 8. NVLab structure.
In this figure, the new tool called Designer
is a simple Java Applet
with a GUI interface that we designed to allow users to draw a network. Its interface ( Fig. 9
) includes a number of buttons for adding devices to the drawing and interconnecting them. The maximum number of devices that can be used depends on the instructor's preferences and can be accordingly changed by the system programmer. For time being, this number is set to 16 which we consider to be suitable for small-scale networks.
Fig. 9. Designer applet.
Designer will ensure that connections made in the network diagram adhere to a rule imposed by the Xen- bridged networking model which allows connections to be made between one VM's interface on one end and an interface on a switch on the other end only.
Through the menu Designer
has ( Fig. 10
), the learners will be able to:
Send the network diagram's information to be saved on the server side as an XML document. ( Send to server.)
Invoke the so-called Builder tool, which will use the XML document, for building the network on the server. ( Create the diagram on the server.)
Clear the drawing ( Discard and Open New Design) and start over from the beginning.
Print the diagram on the learner's local printer. ( Print Current Design on Local Printer.)
When the learner has finished drawing the network diagram and wants to send its information to be saved on the server, he/she can click on Send to Server
option from the menu. This will instruct Designer
to create an XML document containing all information related to the design and initiate an HTTP connection to the remote side over which the document will be sent. The XML document contains information about the devices included in the diagram and the interconnections between them, and is structured as shown in Fig. 11
. This figure has been generated for the simple network drawing shown to the right of the XML document.
Fig. 10. Designer's menu options.
Fig. 11. The XML document created for a simple network diagram.
Selecting the second option in Designer's menu ( create the diagram on the server) invokes the new tool Builder which we designed to take care of creating the devices included in learner's network on the server. Builder will try to load the network diagram XML file the learner has already saved on the server, confirm the success of parsing its content, and then start creating the network by instantiating copies of the prototype VMs and connecting them to the designated software switches.
For a VM to be ready, two files need to be prepared: the storage image file and the configuration file. The storage image file will be exported as a hard disk where the guest OS instance resides. This file will be copied from a default location on the server's file system to the directory where Xen stores all VMs' image files. Copying this file takes around 4-5 minutes depending on the prototype machine image file's size which is determined by the instructor. The configuration file will (as mentioned before) include a description of all resources assigned to the VM to be used by Xen when creating it, and will be generated by Builder. Generating this file does not consume much time since its size is usually within hundreds of bytes.
To test the basic functionality of the system and how useful it can be, a group of 15 learners with different levels of IT backgrounds were asked to try an exercise that is part of Computer Networks course from remote locations.
All of the participants were university students who either had previously studied or attended Computer Networks classes at the time when the exercise was proposed to them. Two individuals had previously been certified as Cisco-Certified Network Associate.
The exercise is based on a sample network comprising two PCs (PC1 and PC2), two software switches (xenbr1 and xenbr2), and a router (R1), as shown in Fig. 12
. This simple network has two broadcast domains and one router device that, when properly configured, can pass traffic between them. It can be used to develop the learner's skills in determining and configuring IP addresses in a network, enabling traffic forwarding between devices connected to different switches, and troubleshooting connectivity problems.
Fig. 12. A case study network design.
Two prototype VMs for a PC and a Router were created in the system, and we selected Linux (Fedora Core 6) to run as guest OS in both of them. A minimum number of packages were installed in order to shorten the VM's creation time. The X window system was not among these packages and the learners were requested to use the command-line prompt to perform each step in the exercise. Furthermore, SELinux and the iptables services were disabled since the task is not designed to provide training on network security concepts and to allow learners to test whatever they may see necessary without having to consider the firewall rules. Learners were able to log into their machines using the "root" account information and this enabled them to have full control over the devices.
To help the learners get used to the system before taking the exercise, we prepared a tutorial with sample movies to illustrate how the different tools work. And because most of the participants had little experience using Linux, we included the commands mostly encountered in networked environments in the tutorial.
The importance of providing static
contents (a text containing explanation of the concepts to be introduced) along with the system tools has been indicated in recent researches [ 25
]. With such an arrangement, learners were able to get more out of what they are presented. Accordingly, we handed a paper that included simple explanation of the targeted skills to the learners to read before or while practicing.
Due to the availability of one server for testing phase, access to the system has been granted to one user at a time, and a period of two hours for each user to complete the task was scheduled. Users were also encouraged to try any idea related to this task that comes into their minds and even try to use more devices if needed.
Evaluating the progress made by taking the exercise was done by introducing one level-test to all participating learners and asking them to complete it prior to starting the task and later after finishing it. This level-test included multiple-choice, yes/no, and fill-in-blank questions on the various concepts included in the exercise. Examples of these questions are shown in Table 1
Table 1. Sample Questions from the Level Test
Furthermore, learners were requested to fill a survey that helps us identify any weak points or improperness of the task/laboratory design. The survey was designed to get general and detailed feedback from learners about what they think of the different tools in NVLab. While the general questions were put to help compare the learners' performance in a similar laboratory environment with the one in NVLab, the detailed fields of the survey queried about the technical aspects of the tools and any difficulties that learners faced using them.
To cover concepts of basic IP planning, connectivity check, and simple routing in Linux machines, a sample task was proposed for the students to complete. Steps to be followed are:
Use the two C class IP subnets 192.168.0.x and 192.168.1.x in this task.
Assign IP addresses from one subnet to the interfaces connected to xenbr1 (eth0 from the router and eth0 from PC1).
Assign IP addresses from the other subnet to the interfaces connected to xenbr2 (eth1 from the router and eth0 from PC2).
Check the connectivity between each PC and the router.
Configure the router to allow IP forwarding and check the connectivity between PC1 and PC2.
6.2 Exercise Result
All participating learners were able to complete the task successfully (snapshots of one learner's machine is provided in Fig. 13
). Comparison of the level-test results for the learners before and after taking the exercise gave two positive indicators as shown in Table 2
Fig. 13. Completed as appeared on the learner side.
Table 2. Number of Correct and Wrong Answers Before and After the Exercise for All Learners
This table shows that the increase in the correct answers was 8 percent and that this was also accompanied by a decrease of 5 percent in the number of wrong answers. The difference between the two percentages arose because learners could select right and wrong answers in the multichoice questions.
Participants also provided their feedback and comments on the survey prepared for this experiment. Analysis of the feedback information contained in the survey which was handed to the learners provided the results shown in Fig. 14
. These can be summarized as follows:
1. On a scale between 0 and 9, where 0 indicates "very bad" and 9 indicates "excellent":
Designer tool rating was in the range of 7-9. Detailed feedback related to this tool showed that some learners found difficulty in drawing the connection between two devices.This is due to the fact that Designer is not intended to perform high-level drawing functions.
Builder rating (6-9) was slightly lower than Designer. In their comments, learners mentioned that they found the time Builder takes for creating the network on the server to be long.
The connection to the server was granted rating in the range of 7-9, but learners commented that the size of the VNC applet window showing the host server's desktop was relatively small which caused inconveniency while taking the exercise.
The size of the VMs' window was also small but had less improper impact on the learners than the impact of the small size of the server's VNC window.
2. On a scale between 0 and 9, where 0 indicates "very difficult" and 9 indicates "extremely easy," most learners found the exercise somewhere between easy and difficult since they were exposed to similar exercise as reflected in the results also ( Fig. 14). Only 20 percent of the sample group found the exercise very easy because they had previous networking experience and further study of advanced networking concepts. Working on Linux machines rather than MS Windows machines made the exercise slightly difficult for most learners.
3. In the exercise text, suggestions to perform more tests on the network, try more than one plan for IP planning, and even try a different network were included. While 60 percent of the learners completed the exercise as handed, 40 percent of them went beyond the exercise text and tried some of the suggestions and other advanced scenarios for the network.
4. All learners who tried NVLab mentioned that they were able to complete the exercise successfully and that they find the system suitable for practicing Computer Networks and that they believe that it will be possible to take more complex exercises within it.
In this paper, we presented our approach toward building a Web-based Computer Networks laboratory (NVLab) where students can freely practice on designing, configuring, and troubleshooting a network scenario.
Fig. 14. Feedback analysis results.
To build NVLab, we used Virtualization and VNC as core technologies. Virtualization enables creating a number of networked VMs on one server, and VNC facilitates remote access to the server hosting these VMs in order to configure them and test their work.
The laboratory introduces two new tools for the learners: Designer and Builder. The learners use Designer for producing a network drawing and sending its information as an XML document to be saved on the server side. Then, they will invoke Builder which is responsible for creating Virtual Machines that correspond to those included in the XML document. In the next step, learners will be able to use their Web browsers to connect to the VNC module installed on the host server, start virt-manager (the application through which they can control the VMs), and then start configuring and testing their network.
Responding to our request, a selected group of students with different levels of networking background tried a sample task covering concepts on IP subnetting, connectivity check, and basic routing within NVLab from remote locations.
All the participants could complete the exercise successfully, and the feedback received from them showed that they were satisfied with the way they worked on their designs. The feedback also indicated some notes on the laboratory interface and we will consider working on them in newer versions of the laboratory system.
Having verified the appropriateness of the system to be used in computer networks course, we are working on:
1. Modifying the laboratory structure to accept more than one user logging to the system.
2. Improving Builder to make the creation of the Virtual Machines faster (currently one VM takes 4-5 minutes to be created).
3. Implementing the laboratory in real environment to test it quantitatively.
M. Wannous is with the Graduate School of Science and Technology, Kumamoto University, Kurokami 2-39-1, Kumamoto-City 860-8555, Japan. E-mail: email@example.com.
H. Nakano is with the Center for Multimedia and Information Technologies, Kumamoto University, Kurokami 2-39-1, Kumamoto-City 860-8555, Japan. E-mail: firstname.lastname@example.org.
Manuscript received 6 Aug. 2008; revised 24 Apr. 2009; accepted 29 July 2009; published online 10 Aug. 2009.
For information on obtaining reprints of this article, please send e-mail to: email@example.com, and reference IEEECS Log Number TLT-2008-08-0063.
Digital Object Identifier no. 10.1109/TLT.2009.31.
received the MEng degree in computer science and electrical engineering in 2009 from Kumamoto University, Japan, where he is currently working toward the PhD degree in the Graduate School of Science and Technology. His main research topic is the design of Web-based ICT laboratories and integrating them into the available courses. He has published a number of papers on this topic in international and Japanese conferences. He is a student member of the IEEE and the IEEE Computer Society and is also a member of the IEEE Education Society.
received the PhD (Doctor of Science) degree from Kyushu University. He is currently a professor in the Center for Multimedia and Information Technology, Kumamoto University. His university teaching experiences are in physics and information science at Nagoya University, and information education and instructional systems (ICT field) at Kumamoto University, where he introduced university-wide e-learning environments and played an important role in founding the Graduated School of Instructional Systems and the Institute of e-Learning Development. His research work focuses on the development of the virtual learning environments, e-laboratory systems and virtual reality for e-learning, and scientific visualization.