CHAPTER FIVE RESULT ANALYSIS 5.1 Chapter Introduction 5.2 Discussion of Results 5.3 Performance Comparisons 5.4 Chapter Summary 61
5.1 Chapter Introduction This chapter outlines the results obtained from implementation of tele-collaborative environment by using AG, Paraview and Paraview plugin. The chapter begins with development of tele-collaboration environment and results obtained. We briefly explain all the steps taken. Other than that, the experimental challenges and performance comparison between some commercialize applications are also discussed. Similar to other chapters, this chapter ends with summary. 5.2 Discussion of Results This section mainly highlights the development results of tele-collaborative environment with AG, Paraview and Paraview plugin. Introduction and experimental environment and results are discussed below. 5.2.1 Experimental Environment In this thesis, the system architecture has four main parts: clients, master client, server and computing nodes. Currently, AccessGrid can support multi platforms such as Windows, Linux and etc, therefore there is no operating system platform limitation for clients. At client node, clients only need to have AG 3 and AG-Paraview plugin. A quad core processors computer is being used as computing nodes and server. In this computer, VTK, Paraview and openmpi and other related applications are being installed (eg: Java, Python and etc). Other than that, y1 computer had been used as the master client where Paraview, openmpi, AG, AG-Paraview plugin and Paraview Plugin are installed. The hardware specifications for server and computing node are shown in Table 5.1 62
Table 5.1 Hardware Specification for Computing node and Master Client Hostname Quadcore y1 Operating System Linux opensuse Microsoft Windows CPU Intel Xeron X3220 Intel Core 2 Duo RAM 2.40GHz 1GB 1.66GHz 2.99GB No. of Processors 4 2 5.2.2 Experimental Results In this research, the quadcore computer worked as server and computing nodes. In this case, data server nodes and render server nodes are sharing the same process. However, clients are run separately. Paraview is compiled with MPI, four logical nodes are being created in the server. >> mpirun np 4./pvserver >> Listen on port: 11111 >> Waiting for client... Master client (y1) connects to server by adding new server into Paraview. Configuration setting must be set before start. The created new server screenshot is shown in Figure 5.1. Once master client is connected to the server, the screen will show Client connected. >> mpirun np 4./pvserver >> Listen on port: 11111 >> Waiting for client... >> Client connected. 63
Figure 5.1: Add New Server Until current stage, a client/server parallel launch mode had been developed. However, there still few more work to do to allow shared visualization among other clients. Therefore, master client is plugged in a Paraview streaming plugin to allow result rendered in Paraview to be streamed to remote receivers as video. The remote receivers (clients) could be able to see Paraview master client s interactions with Paraview in near real-time (25 frames per second) at native resolution and need not have Paraview installed. Figure 5.2 shows the streaming plugin of Paraview. Master client can set a configuration to allow streamed visualization by launching vic against the multicast address. Eg: vic 224.2.1.15/22222 Figure 5.2: Streaming Plugin in Paraview 64
At clients side, clients launch AG Venue Client. AG users must have AG-Paraview plugin before entering to any AG Venue as shown in Figure 5.3. In this research, the default venue address used is https://vv3.mcs.anl.gov:8000/venues/default where the default services had been fully provided; all clients are connected using multicasting mode. All clients must have the same setting of VIC and RAT as follow: RAT: H.261, Frame Rate 24 f/s, Rate Control 128 kb/s VIC: Sample Rate 8 khz, Bit Rate 16kb/s Figure 5.3: Streaming plugin in AG In Master Client (y1) manipulation of heart dataset are allowed. Volume slicing and rendering can be easily carried out. Results of rendering, slicing and others will be rendered to clients through VIC streaming nearly in real-time. However, the result shown at client s nodes is lower quality 2D visualization. Master client is able to enable audio control for discussion and video display and capture. In Figure 5.4 shows that 3D data in Master Clients. Every participant can be communicated through Video Devices and RAT. In Figure 5.5, the client can viewed the 2D output from the Master Client although no Paraview application is installed. The client also can communicate with Master Client with AG services. 65
Figure 5.4: Screenshot of Master Client Figure 5.5: Screenshot of one of the client 66
5.2.3 Results Analysis In Table 5.2, the results gathered are been analysed. When adding more clients, quality of audio will getting noisier and a lot of echo waves will affect hearing and sometime we cannot differentiate on owner of voice. As per the result shown, the visualized results are in 2D and not in 3D. Point clouds originally manipulated in master client can be clearly visualized (Figure 5.6), while the results viewed by clients through VIC streaming, the points cannot clearly be visualized (Figure 5.7). In Figure 5.8, the heart dataset is compared with time taken to rendering in Master Client, and also the time taken for the results being streamed to clients. According to the results, the larger scale of dataset will also increase the time to render. However, the size of point clouds gives small effect to time of streaming the result from master client to clients. This result shows that the output of visualization will not be affected by the size of point clouds. Several manipulations can be carried out by Master Client. Figure 5.9 shows the result of extraction on 3D heart dataset with selection by using Extract Geometry Filter in Paraview. Figure 5.10 shows the vtkpvconnectivityfilter applied to the 3D heart model. This filter will extract cells that meet connectivity criteria. The filter output the largest connected region in the dataset. 67
Figure 5.6: Original heart dataset from top at Master Client Figure 5.7: Output visualized at client side 68
Rendering of Point Clouds Time (seconds) 16 14 12 10 8 6 4 2 0 Rendered Time at Master Client Time rendered from master to client 5 0 0 0 0 6 0 0 0 0 7 0 0 0 0 8 0 0 0 0 9 0 0 0 0 1 0 0 0 0 0 Heart Datasets (data points) Figure 5.8: The rendered time taken at Master Client and streamed out time taken to visualized at Client Side with various point clouds Figure 5.9: Output of Extract Selection Filter 69
Figure 5.10: Output of Connectivity Filter Table 5.2: Analysis Results Gathered Criteria Master Client Client Quality of Output Clear and 3D Blur of output and 2D Quality of Audio Noises when more clients Same as Master Client entering the session. Feature Parallel Processing can be carried out. Various types of visualization methods can Only result of visualized results can be viewed. Can not change input of dataset and visualization. be applied using Paraview. Voice and text Voice and text conversation allowed conversation allowed through AG through AG Large scale point clouds Parallelism Affected on time rendering Parallelism processing will reduce the time of rendering in master client. With higher computing nodes, the rendering time in master client will be reduced. Small affected on time from master client to share among clients. Not much effects on the time of streamed from master client to clients. 70
5.3 Performance Comparisons 5.3.1 Comparison with Instant Software MSN and Skype Currently widely used instant messaging software application such as MSN and Skype is a type of simplified video conferencing tool which allow multiple person to communicate with each other. However, most of them only support point to point conferencing. Other than that, some other shared applications such as shared browser, presentation slide are not available in those applications. Basically, they only support instant messaging between two persons and file transfer. Table 5.3 shows the comparison between AG and MSN, Skype. Support Video Enlarge Group Size for video conferencing Table 5.3: Comparison among AG, MSN and Skype AG MSN Skype Yes No No Support more than 20 points Point to point Point to point (with extended supported software can support 6 points) Support Multiple Yes No No video streamed out Network Multicast/Unicast Point to point Point to point Connection Bandwidth for 16K (Support multiple 6.4K 100K messaging codec) Extendable Extended with other No No software such as VTK and Paraview for 3D Visualization Support Text Chat Yes Yes Yes Support File Yes Yes No Sharing Operating Complicated to operate Easier to operate Easier to operate Support Other Shared Application No No Other shared application such as shared presentation, shared browser and etc 71
5.3.2 Traditional 3D Visualization vs Tele-collaborative 3D Visualization Traditionally, a lot of conferencing tool had been developed to support normal meeting which allow company staffs or institute managers to communicate with each other although they are geographically separated. However, most of these tools required high resources support tool such as Multiple Control Unit (MCU). Therefore, a lot of 3D Visualization had been limited by the hardware requirement and will still require a face-to-face meeting for results discussion which is costly and time consuming. AG is a freeware which support multicasting connection and also extendable to support 3D visualization images from Paraview. 5.3.3 Three Dimensional Heart Data Visualization Comparison The visualization of the heart done using the IBM was also compared with the visualization done by Kounchev and Wilson in [40]. Kounchev and Wilson visualized the left ventricle of the heart surface. In our research we are focused on the complete heart model. Figure 5.11 show our heart model visualization with Kounchev and Wilson s. With the visualization of full heart model will allow more understanding on heart study. 72
Figure 5.11: Comparison of Heart Visualization 5.4 Chapter Summary This chapter highlights the results and analysis done to approaches implemented in Chapter 4. The result gathered show that a tele-collaborative visualization system is successfully developed for 3D heart model. The system allowed sharing the rendered results of 3D heart model with Paraview through AG with nearly real time. The quality of result is not as high quality as the original rendered data at master client. The large scale dataset will not highly affect the time streamed results from master client to all clients. The comparison between other similar low cost conferencing tools such as MSN and Skype show that AG can support more people communication and extendable for 3D Visualization Tool (Paraview). The full set of 3D heart model will also give a clearer heart physical functionality for heart study. 73