Two technical factors to migrate GIS into web environment

Distributed Processing Model for Geographical Data Reference

Figure 3. New Distributed Processing Model for Geographical Data Reference

As discussed in a former section, conventional Web GIS model is implemented widely on general Web applications, because this model is independent from computer capacity and platform type of Web client and is able to distribute geographical data widely. However, data process is concentrated on GIS servers, causing a serious bottleneck which affects total system performance.

In these days, capacity of personal computers bears comparison with that of servers except for reliability. We have invented a new model for decreasing overhead of GIS servers by fully utilizing PC clients capacities which is more than enough just for Web browser manipulation. In other words, we transfer data manipulation process from GIS server to PC clients because PC clients hold enough resource to do that job. Figure 3 illustrates the procedure.

The distinctive point of this model is to transfer spatial index files which contain a structure and actual address of geographical data on data site to PC clients when they initially access a Web server. At this point, PC clients are informed how to access actual geographical data on data sites, so each PC client simply requires a Web server to load specific geographical data files by referring to spatial index files which describe types and areas of each geographical data.

Loaded geographical data is displayed on PC clients via Web browser process. Thus no transaction process exists on a GIS server.

That is, only WWW server operates on a Web server and serves ordinary HTML homepage distribution process. Since sequential data conversion process on a GIS server is eliminated, multiple clients requests are processed concurrently and required geographical data are distributed all at once. Demonstration Web GIS site is implemented on our company’s homepage (http://www.dawn-corp.co.jp).

Figure 4. Concurrent Data Distribution by Distributed Processing Model

Spatial Index Files
As discussed in a former section, spatial index files support distributed processing models. Figure 5 shows the outline of spatial index files.

A spatial index file represents a configuration list to show how a group of geographical data files is positioned. We have developed a kind of memory caching mechanism to load only a required area of geographical data files onto memory by referring to this spatial index files. This mechanism allows us to manipulate enormous volume of geographical data as a seamless virtual map to be displayed quickly on demand. Moreover, the capability to configure spatial index files hierarchically enable us to switch different scales and types of maps instantly.

Figure 5. Spatial Index File

High Performance and High Functionality
Geographical data loaded onto memory of PC clients can be displayed and manipulated quickly by utilizing clients resources. Basically, loaded geographical data are stored on memory of PC clients until memory overflow occurs, thus data reload overhead is minimized. In addition, high function spatial analysis such as polygon logical operation and border line search can be operated only on PC clients. In implementing GIS in response to market requirments, special functions are required according to each type of business. In case of implementing server/client system, the more function you require, the higher cost you have to bear. You have to define communication protocol according to each man-machine interface and develop programs for both server and client sides. In implementing the same system on distributed Web environment, most of the function can be implemented only on PC clients, leading to simple and low-cost implementation. Of course some functions such as wide area network search are best fitted to be processed on a server.