y for 
e and 
>s the 
e the 
1.0.0 
ilable 
1. On 
than 
users 
0 the 
When 
oding 
coded 
e and 
quest 
in the 
build 
y and 
1g the 
em(Li 
ching 
incing 
'esults 
tem is 
chine, 
‘the 4 
tching 
equest 
ervice 
ash as 
'd to a 
:tively 
system 
it can 
ervice 
r. The 
ion is 
  
      
GeoNetwork Ux ^ GeoNetwork 
GeoNetwork 
Figure2.Web cluster 
2.Disaster Recovery 
Disaster recovery here means that the metadata service system 
crashed or can't serve normal, the natural disasters, 
infrastructure failure are not involved. GeoNetwork uses JDBC 
to operate database, when the number of metadata is large in 
one operation, memory overflow may be happen, it can leads 
to system crash or can't respond to requests; user concurrent 
access may also lead to system can't support, it is needed to 
design an appropriate program to help the system return to 
normal state as soon as possible. 
The metadata service system uses monitoring keywords for 
restarting service method to restore. GeoNetwork uses 
Wrapper to install as a Windows service, we can use filter 
mechanism on Wrapper. In the filter, we can use monitoring 
keywords as trigger string, like "RESTEART NOW”, and the 
trigger action can set to service restart. After the system 
running, we can throw the keywords when we need, it can be 
monitored by Wrapper, and then Wrapper can restart the 
system. Our system can throw the keywords when catch the 
memory overflow exception, and Wrapper monitors the string, 
trigger the filter, and act to restart the service system. This 
method can also be used for service remote management, like 
restart to apply new settings. 
Actually previously mentioned *4--1" model is also a kind of 
disaster recovery scene. When any one of the 4 normal server 
crashed, the front-end dispatcher can monitor and dispatch new 
request to the 1 backup server, so the cluster can remain stable 
service capability. 
3. RESULTS 
An experiment on Dell Precision T3400( OS: Windows XP sp3, 
JVM parameter “-Xms48m —Xmx1024m”) has been done to 
verify the effectiveness of HOM-improved solution. The system 
efficiency is compared in Table 1 and Figure 3. 
The header of the table is the volume of metadata, and the unit 
is seconds, which means the computer need the time to finish 
the metadata volume. 
  
Software\volume | 100(S) | 1000(S) | 10000(S) | 100000(S) 
  
GeoNetwork2.1 5.62 68.074 2313.427 X 
  
  
  
  
  
HOM-improved | 3.692 31.182 300.411 2980.333 
  
  
X: can not be imported one time on the amount level 
Table 1. Import efficiency contrast on different amount of 
metadata 
  
  
Fa HO M-improved 
  
  
  
  
  
2804 —^-— Geghetwork2. 1 
tt. 
A A 
£ ZU 
200 - os 
© 
et 
© a 
es ^ 
S 450-4 s 
© Ps 
E T 
5 e 
e 16€ - t 
o JL 
m d 
= Lu 
= osi " 
T + T * T Y T T * nn, 
Q 2000 4000 £006 £000 10900 
metadata volume 
Figure 3. Time per metadata contrast on data importing 
From Table 1, we can see the efficiency of batch import 
function through HOM-improved software compared to the 
original GeoNetwork 2.1 can increase 2-10 times, and the 
greater the amount of metadata, the higher the efficiency. 
From Figure 3, in GeoNetwork 2.1 metadata batch import 
function has time consumption growth rate is far greater than 
the amount of data, while the HOM-improved on the contrary, 
the time consumption rate is less than the amount of data about 
growth rate. 
In summary, HOM-improved solution will be more responsive 
to the amount increasing of metadata amount, it is adaptive to 
our metadata service system. 
4. CONCLUSIONS 
GeoNetwork as a geographical spatial metadata service, can 
used to publish, search metadata, is the base software for our 
metadata service system. The Hierarchical Optimization Model 
has been presented for preventing the original GeoNetwork 2.1 
shortcomings when serving as an internet application. Based on 
the HOM-impoved solution, we break through the bottlenecks, 
efficiently improve the function efficiency , load capacity and 
the system performance. Next we will submit our model and 
source code to GeoNetwork project. 
REFERENCES 
Jeroen Tichler, Jelle U. Hielkema, 2007. GeoNetwork 
opensource Internationally Standardized Distributed Spatial 
Information Management[J].OSGeo Journal.2 
Gong Jianya, Du Daosheng, Gao Wenxiu, Xu Feng, Zhou 
Xu, 2009. Technology and Standards of Geographic 
Information Sharing[M]. Beijing: Science Press. 
JIN Zhi-guo, SHOU Chun-fa, LI Cheng-ming, YIN jie,2008. A 
discussion of the mode of urban geoinformation distribution 
service based on network [J]. Science of Surveying and 
Mapping.33(6).pp.196-198 
HE Chen, CHEN Zhao-xiong, HUANG He-yan, 2004. 
Summary of Web Caching Technology [J]. MINI-MICRO 
SYSTEMS. 25(5).pp.836-842