ISPRS Workshop on Service and Application of Spatial Data Infrastructure, XXXVI(4/W6), Oct. 14-16, Hangzhou, China 
geographic information service and provides a standard 
interface for users. By calling the interface, users can obtain 
geographic information service. Each GIS component that is 
hosted on GIS application server executes a geographic 
information service function, such as map searching, terrain 
analysis and so on. 
Component container, which is a system layer GIS component 
based on, provides space, running environment and system 
service for GIS component. It is also responsible for the 
communication between users and GIS components. All the 
GIS components run on component container. And as the 
system service is provided in system layer, developers don’t 
need any consideration about them. And what they need to do is 
the geographic information service logic designing. 
Component management unit provides a man-machine interface 
for developers to facilitate the GIS components management, 
including GIS component building, installation, startup, stop 
and deleting. 
Connection management handles the connection with GIS 
database. It also maintains a database connection pool which 
used to release system burdens that will be taken place when 
multi-users’ concurrent requests posted. 
The public service, which is correlated to system layer, defines 
a series of programming interface that is used by GIS 
component. It is different from geographic information service 
which is correlated to application layer. The two kinds of 
service are totally different. Public service is the public portion 
shared by all GIS components. It provides system service for 
GIS component. And, geographic information service serves for 
users and provides them with GIS functions. 
5. CHARACTERISTICS OF THE MODEL 
5.1 Reliability 
In this model, geographic information service is implemented 
by GIS component. Each of the components realizes a GIS 
function. They are relatively independent and do not affect each 
other, so the system still can provide geographic information 
service to users even though there is something wrong with one 
component. In addition, by adding more GIS application server 
to the system and clustering them, if there is a server fail to 
work, we can transfer some user requests to other machines. 
The system is still working. 
5.2 Expansibility 
In this model, component and framework technology was 
adopted. Therefore, we can adjust geographic information 
service by adding or deleting GIS component according to the 
needs of users. This makes the model with a high expandability. 
When a new geographic information service is needed, 
developers can meet the need just to add a GIS component that 
implements the service to the model. Furthermore, we can 
enlarge the scale of the model—by adding more GIS application 
server and clustering them—in case of the system insufficient 
when multi-users access system simultaneously. 
5.3 Development Efficiency 
The model introduces modularized development method, and 
divides geographic information service into GIS component. So, 
for developers, they don’t need to care about the technique 
details of the system layer but focus on component designing. 
Since the components in this model are reuse modules, 
developers can build new service by assemble exiting 
components into a new component that realizes the new service. 
So the model provides an efficient way for developers to 
develop a new system. 
6. KEY TECHNOLOGIES 
6.1 Load Balance 
Because of the heavy data processing on geographic 
information analysis, load balance technology is very important 
to GIS application server. It balances loads on every computer 
by distributing them reasonably, and avoiding some computer 
over load but others under load. 
To realize load balance of the model, load assignment algorithm 
and the service implement techniques can be employed. At 
present, there are two algorithms available for load balance. 
One is the precise load assignment algorithm. In this method, 
there is a distributor. When a request is received, system first 
inquires distributor. Then, according to the algorithm, the 
distributor decides which application servers process the request. 
The other is statistic load assignment algorithm. In this method, 
the requests will be assigned to a server according to the 
probability and its weight which has been set in advance. 
Another way to balance load is by using different service 
implement techniques. One technique is that service 
implemented based on process. Application server creates all 
needed process initially, when a request is received, it will be 
handled by a process that already exits. So, the system does not 
create and terminate process frequently. The other technique is 
that service implemented based on thread. Application server 
only creates necessary process, when a request is received, 
system creates a thread temporally and hand the request to the 
thread. The thread will be killed after the response of the 
request returned. 
6.2 Cache Memory 
Cache memory is used to save response of user’s request. It is 
very essential for the model because it greatly improves system's 
speed and efficiency. When a request, which had been 
processed before, is received, the system just returns the 
response that has been saved in cache directly to user but 
doesn’t need to call GIS component. Thus, the system response 
time is highly decreased. According to the contents saved in 
cache, we can apply cache memory technology to the model in 
two ways. One is to apply the technology in application layer, 
which creates cache to store web pages. The other is to apply it 
to data layer, which creates cache to store datum from database. 
6.3 Database Connection Pool 
Connecting to a database is a time consuming activity since the 
database must allocate communication and memory resources as 
well as authenticate the user and set up the corresponding 
security context. Establishing the connection once and then use