Efficient Shortest Path Routing Algorithms for Distributed XML
Processing
Ye Longjian
1
, Hiroshi Koide
2
, Dirceu Cavendish
3
and Kouichi Sakurai
4
1
Graduate School of Information Science and Electrical Engineering, Kyushu University, 744 Motooka,
Nishi-ku, Fukuoka, Japan
2
Research Instiute for Information Technology, Kyushu University, 744 Motooka, Nishi-ku, Fukuoka, Japan
3
Network Design Research Centre, Kyushu Institute of Technology, 680-4 Kawazu, Iizuka, Fukuoka, Japan
4
Faculty of Information Science and Electrical Engineering, Kyushu University, 744 Motooka, Nishi-ku, Fukuoka, Japan
Keywords: Distributed XML Processing, Web Service Processing, Web Service Processing Offload.
Abstract:
This paper analyses the problem of efficiently routing XML documents on a network whose nodes are capable
of distributed XML processing. The goal of our study is to find network paths for which XML documents’
transmission will result in high likelihood that a large portion of the documents be processed within the net-
work, decreasing the amount of XML processing at documents arrival at the destination site. We propose
several routing algorithms for single route and multipath routing and evaluate them on a distributed XML
network simulation environment. We show the benefits of the proposed XML routing algorithms as compared
with widespread minimum hop routing strategy of the Internet.
1 INTRODUCTION
Web services has become an indispensable infras-
tructure for our society. A multitude of Web ser-
vices are currently deployed, demanding significant
resources, such as CPU power and memory space,
to support their services quality. The processing of
Web services are generally provided at end points
only in the current Web services. With widespread
usage of cloud processing, we envision a distributed
Web service processing approach, where part of Web
server’s processing can be offloaded to intermediate
nodes in network, reducing end points load and im-
proving services throughput. This Web service net-
work processing paradigm may lead to efficient re-
source usage with effective scheduling and higher
quality of services. Currently, XML data is one of
the basic communication formats in Web services in-
frastructure. Servers typically process XML data in
various Web services (e.g. collaborative services).
(Cavendish and Candan, 2008) has proposed a dis-
tributed XML processing platform, using intermedi-
ate network nodes besides clients and servers. They
have shown from an algorithmic point of view how
to turn well-formedness, grammar validation, and fil-
tering into distributed processing tasks. In a subse-
quent work, (Y. Uratani and Oie, 2012) has studied
distributed XML processing performance of network
nodes capable of XML document processing, for var-
ious network topologies and document types.
This paper addressed the problem of routing XML
documents across a network where nodes have vari-
ous XML processing capabilities. Our objective is to
find routes at which a large amount of XML process-
ing be executed by network nodes, saving process-
ing efforts at destination nodes. With the objective
of increasing XML processing at network nodes, we
tackle the problem of how best route XML traffic, and
introduce single path available capacity (ACAP) and
latency available capacity routing (LACAP), as well
as multipath N-route shortest path and N-route dis-
jointed shortest path XML routing algorithms.
The paper is organized as follows. In section 2,
we address related work. In section 3, we describe
a distributed XML processing network environment,
where nodes are able to execute basic XML pro-
cessing tasks, such as well-formedness checking and
grammar validation. We also define the XML routing
problem, and introduce various routing algorithms. In
section 4, we describe a distributed XML process-
ing experimental environment, and characterize XML
routing performance. In section 5, we summarize our
findings and address future research directions.
Longjian, Y., Koide, H., Cavendish, D. and Sakurai, K.
Efficient Shortest Path Routing Algorithms for Distributed XML Processing.
DOI: 10.5220/0008162002650272
In Proceedings of the 15th International Conference on Web Information Systems and Technologies (WEBIST 2019), pages 265-272
ISBN: 978-989-758-386-5
Copyright
c
2019 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
265
2 RELATED WORK
We assume a distributed XML processing paradigm,
where intermediate nodes are capable of XML pro-
cessing. Recent support for this paradigm is as fol-
lows. Active Network (Tennenhouse and Wether-
all, 2007) also focuses on processing at intermediate
switches. The processing manner is assumed to be of
two types: type-1) encapsulated into a packet, type-2)
assigned to network switches beforehand. The type-
1 manner executes only simple processing but is able
to execute fast processing because of hardware execu-
tion. The type-2 manner, which is more similar to our
system, is able to process more complex XML tasks.
VNode (Y. Kanada and Nakao, 2012) also provides
a processing environment at intermediate switches.
In their research, the processing function is also pro-
vided at customized switches and its processing envi-
ronment is provided as a virtualized environment us-
ing a virtual machine. These researches provide not
only transport functions but also processing functions
in the network.
Next we show current researches of specific pro-
cessing in networks. In transcoding (S. H. Kim and
Ro, 2012), a content server delivers data (e.g. video
data) to clients via a transcoding server. The transcod-
ing server transforms the original data to data which
reflects user’s demands. For instance, the data may be
transformed from high resolution to low resolution at
the transcoding server to adapt to mobile devices. An-
other intermediate node processing is a cache server
(S. Nishimura and Ikenaga, 2012; Kalarani and Uma,
2013), a key technology of content delivery net-
works. Cache servers are allocated to wide distributed
places and store contents as cache from other con-
tent servers, driven by user request patterns. Upon
user’s request, data is delivered from the nearest cache
servers leading to lower network latency. (Fan and
Chen, 2012; Solis and Obraczka, 2006) focus on sen-
sor networks. These researches propose to consol-
idate the large amount of sensing data at some in-
termediate nodes before large data reach data col-
lection servers. Such approach can reduce energy
consumption and network load for mobile sensor de-
vices. (M. Shimamura and Tsuru, 2010) studies the
compression of packets near a sender, expanding the
packets near a receiver during buffer queueing time to
achieve better network resource utilization. In these
works, we see that the network provides special func-
tions such as video transformation, data caching and
so on for specific services.
Finally, few papers have addressed network rout-
ing problem for efficient XML processing. (Ziyaeva
and Min, 2008) addresses the problem of routing
XML content to appropriate recipients within an En-
terprise Service Bus (ESB), where specific XML pro-
cessing is required. However, in contrast to our work,
XML processing is executed only at the recipient’s
site. In (Wang and Ozsu, 2007), the authors address
the problem of routing XML queries in an efficient
way over a large Peer-to-Peer network. The objec-
tive there is to best satisfy XML queries at destination
nodes, as opposed to XML execution at intermediate
nodes.
3 DISTRIBUTED XML
PROCESSING
Distributed XML processing requires some basic
functions to be supported:
Document Partition: The XML document is di-
vided into fragments, to be processed at process-
ing nodes.
Document Annotation: Each document frag-
ment is annotated with current processing status
upon leaving a processing node.
Document Merging: Document fragments are
merged so as to preserve the original document
structure.
XML processing nodes support some of these tasks,
according to their role in the distributed XML sys-
tem. XML document processing involves stack data
structures for tag processing. When a node reads a
start tag, it pushes the tag name into a stack. When a
node reads an end tag, it pops a top element from the
stack, and compares the end tag name with the popped
tag name. If both tag names are the same, the tags
match. The XML document is well-formed when all
tags match. In addition, in validation checking, each
node executing grammar validation reads DTD files,
and generates grammar rules for validation checking.
Each node processes validation and well-formedness
at the same time, comparing the popped/pushed tags
against grammar rules. Details of these node dis-
tributed processing is described in (Cavendish and
Candan, 2008; Y. Uratani and Oie, 2012).
3.1 XML Routing
An overlay XML network presents networking nodes
with various XML processing capacities. One way
to capture node processing capabilities is to define a
node processing capacity of X XML tags per second
(C = X/sec). That being the case, a routing problem
may be defined as follows. For each XML document
WEBIST 2019 - 15th International Conference on Web Information Systems and Technologies
266
to be transported across the network between source S
and destination D, we wish to find the route such that
the XML document gets the most XML processing
at the network, saving destination D processing re-
sources. Obviously, a favorable route is the one which
contains many XML processing nodes with high pro-
cessing capacity.
The majority of routing algorithms in the Inter-
net use a so called minimum-hop scheme, in which
a path of minimum number of network hops is com-
puted. The rationale is that a min-hop path consumer
less network resources to transport the data. Being
“blind” to XML processing capabilities of network
nodes, this routing scheme may be detrimental to our
goal of maximizing XML processing in the network.
3.2 Minimum Hop Routing and XML
Figure 1 illustrates how a minimum hop routing may
prevent XML processing offloading. In the network,
only node 3 has XML processing capabilities. A min-
imum hop routing algorithm may route XML docu-
ments from node 0 to node 4 via route 0-1-4 only,
with no network node capable of XML processing,
whereas route 0-2-3-4, with three hop count, could be
used for XML processing at node 3.
Figure 1: Example of Network for Minimum Hop Routing.
3.3 Dijkstra XML Routing Algorithm
The first XML routing scheme we propose is as fol-
lows. Lets define a link weight cost as:
w(v
i
,v
j
) =
MaxXMLCap
1 +C
v
j
(1)
where MaxXMLCap is the highest XML processing
capacity among all network nodes, while C
v
j
is the
processing speed of each intermediate node v
j
. With
this definition, if node v
2
has XML processing capac-
ity of 0, all network links arriving at v
2
will have
link cost of MaxXMLCap. For nodes with higher
XML capacities, their incoming link costs are less
than MaxXMLCap. A Dijkstra routing algorithm
(Cavendish and Gerla, 1998) can then be used to find
the path of minimum cost, where the cost PC
p
of a
path p is defined as the sum of the costs of its links,
or:
PC
p
= Σw(v
i
,v
j
) w(v
i
,v
j
) p (2)
3.4 Single Path Routing Algorithm
3.4.1 Capacity Routing
Link cost definition 1 is not practical because it re-
quires the computation of the maximum XML capac-
ity among all XML processing capable nodes. A more
convenient way to define link cost weight is as fol-
lows:
w(v
i
,v
j
) =
1
1 +C
v
j
(3)
In this equation, w(v
i
,v
j
) is the weight of node v
j
. For
the nodes without any capability of processing, C
v
j
is
0, resulting on a weight of 1, which is larger than the
weights of XML processing capable nodes. We call a
Dijkstra routing algorithm using this weight Capacity
Routing.
3.4.2 Available Capacity Routing
Equation 3 does not take into account current XML
processing load of a node, which could lead to pro-
cessing overload. To take into account processing
load, we adjust weight computation as follows. new
calculation method called Available Capacity Routing
whose weights can change over time. This calculation
method is defined as follows:
w(v
i
,v
j
) =
1
1 +C
v
j
An
(4)
An is an available capacity parameter, 0 An 1. As
such, a heavy loaded node will result in higher weight
than a lightly loaded node, causing the routing algo-
rithm to select routes around the heavy loaded node..
3.4.3 Latency Available Capacity Routing
Network latency may affect the quality of the trans-
port service. In this case, an efficient routing strat-
egy seeks to minimize the total network latency, rather
than minimum hop count. If we define l(v
i
,v
j
) as the
link latency between nodes v
i
and v
j
, then the total
latency of a path PC
p
becomes:
PL
p
= Σl(v
i
,v
j
) l(v
i
,v
j
) p (5)
Efficient Shortest Path Routing Algorithms for Distributed XML Processing
267
However, in our XML routing problem, there is
a tradeoff between routes with low path latency and
high XML network processing. This is because low
latency paths correlate highly with low number of
XML processing capable nodes. In other words, a low
latency routing strategy may result in poor network
processing performance. That being the case, a rout-
ing strategy that favors high latency paths may spread
the XML processing around more XML processing
capable nodes, improving performance. Notice that
this strategy will impact end-to-end network latency,
but we argue that Web Service applications tolerate a
reasonable amount of extra latency, differently from
real-time applications such as networked interactive
gaming. Therefore, we adjust the link weight compu-
tation as follows.
w(v
i
,v
j
) =
1
1 +C
v
j
An l(v
i
,v
j
)
(6)
Notice that as l(v
i
,v
j
) gets larger, the weight de-
creases. A limitation of this weight definition is that
if two nodes have two parallel links, the path includ-
ing the longest link is selected during path computa-
tion. However, in this corner case node v
i
may in-
dependently send traffic via the shorter link to v
j
, if
link bandwidth is available. In high speed networks
of today, link bandwidth may be considered infinite
for XML processing, as the amount of XML data is
insignificant as compared to link capacities of tens of
gigabits/sec. This is also the reason that we do not use
bandwidth as part of a routing strategy in our work.
3.5 Multiple Path Routing Algorithm
We have introduced three weight calculation meth-
ods:CAP (Capacity Routing), ACAP (AvailableCa-
pacity Routing) and LACAP (Latency Available Ca-
pacity Routing) which are used in single routing al-
gorithm. To make best use of XML processing capac-
ity of intermediate nodes, we propose multiple path
routing as follows. We have two types of multiple
path routing algorithms: N-route Shortest Path Rout-
ing which uses Dijkstra’s algorithm to compute not
only the shortest path but also second, third shortest,
to as many paths as desired; N-route disjointed short-
est path routing which is the same as previous multi-
path routing except that it avoids repeated usage of a
node in different paths. These two multiple path algo-
rithms will be introduced detailed in 3.5.1 and 3.5.2.
3.5.1 N-route Shortest Path Routing
To calculate the shortest, second shortest, and so on,
we improve the Dijkstra’s algorithm with the follow-
ing steps. Without loss of generality, let a network
have n nodes, with start node 1 and end node n.
Step 1: Calculate all the shortest distances from the
start node to each intermediate node with Dijkstra
algorithm. We get a sequence of n 2 elements
from S(1,2) to S(1,n 1).
Step 2: Calculate all the shortest distances from each
intermediate node to the end node with Dijkstra
algorithm. We get a sequence of n 2 elements
from D(2,n) to D(n 1,n).
Step 3: Shortest path L(t) passing through interme-
diate node t will be : L(t) = S(1,t) + D(t,n).
L(2),L(3),...,L(n 1), when the sequence is ar-
ranged in ascending order, it will result in the
shortest path, the second shortest path, the third
shortest path and so on.
We use N-route shortest path to transport an XML
document efficiently.
3.5.2 N-route Disjointed Shortest Path Routing
In N-route shortest path routing proposed, one node
may be used repeatedly in different routes. To avoid
repeated use of a node, we introduce N-route Dis-
jointed Shortest Path Routing as follows:
Step 1: Calculate the shortest distances from the start
node to target nodes.
Step 2: Delete the nodes from the topology which
are capable of XML processing in the shortest
route.
Step 3: Calculate the shortest distances from the start
node to target nodes as the second route.
Step 4: Repeat the step 2 and 3 to get as many paths
as desired.
We use N-route shortest paths to transport XML
documents while processing them in the network
nodes.
4 EVALUATION EXPERIMENT
FOR XML ROUTING
4.1 Experimental Environment
To evaluate the various XML routing algorithms, we
use a virtual simulation environment, written in the
Scala language. The simulated topology is shown
in Figure 2. The shaded nodes are XML processing
nodes, node 0 is the start node or source, and node 8
is the terminal node or sink. The number between two
WEBIST 2019 - 15th International Conference on Web Information Systems and Technologies
268
nodes represents the distance between them: 1000
distance corresponds to a latency of 10ms. We vary
the XML document size (number of tags), document
generation frequency and node processing capacity.
The minimize hop routing algorithm (MIN) is also
simulated to compare its performance with the XML
routing algorithms introduced.
Figure 2: Distributed XML Network Simulation Topology.
4.2 Experimental Results
In all figures in this session, the Y axis repre-
sents the network processing ratio (network processed
tags/total tags) of the system, the X axis is the tag
generation speed (tag/s). The title of each sub-graph,
such as “document size = 10”, denotes the XML doc-
ument size in number of tags. The labels (min10,
cap10, acap10 ...) represent the processing speed of
intermediate nodes which are capable of processing
and the type of single path computation used: cap10
means capacity routing, and so on. In total, we simu-
lated 36 instances of single path XML routing, 36 in-
stances of two path routing, 36 instances of three path
routing, and 36 instances of two disjoint path routing.
4.2.1 Single Path Routing
In Figures 3, we first notice that for a same routing
algorithm, higher capacity processing nodes lead to
higher network processing ratios, as expected, for all
document sizes. When comparing different routing
algorithms, we find from higher to lower process-
ing ratios LACAP > ACAP > CAP > MIN. The re-
sults verify that in fact including latency on the rout-
ing computation as proposed spreads the XML traf-
fic to more XML processing capable network nodes,
delivering best processing ratio performance. Also,
we verify that minimum hop routing results in worst
performance, because the routes selected use a small
number of XML processing capable nodes, and con-
centrate traffic on the same nodes.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(a) document size = 10.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(b) document size = 50.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(c) document size = 100.
Figure 3: One route shortest path routing.
4.2.2 Multiple Path Routing
Figures 4 and 5, report on two path XML routing and
three path XML routing algorithm results for various
document sizes. For small document sizes (Figs. 3
a, b, and Figs. 4 a, b), we see that processing ra-
tios are similar computation strategies with match-
ing weights. For a large document size, 100 tags,
we see that in general a two path routing delivers
Efficient Shortest Path Routing Algorithms for Distributed XML Processing
269
a higher network processing ratio than single route.
When comparing two route and three route results,
there is no significant improvement for using one ex-
tra route, suggesting a saturation point beyond which
more routes do not lead to better network XML pro-
cessing performance.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(a) document size = 10.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(b) document size = 50.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(c) document size = 100.
Figure 4: Two route shortest path routing.
When comparing two path routing with two dis-
joint path routing (Figs.4 and 6), we can see that
two disjoint path routing performs slightly worse than
two path with node repetition, especially for ACAP
and LACAP weight computation. This is because the
weight computation already takes care of previously
routed traffic load, so repeating a node will not cause
overload processing at a given node. For CAP and
MIN hop weight computation strategies, multiple dis-
joint paths may deliver higher network XML process-
ing performance.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(a) document size = 10.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(b) document size = 50.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(c) document size = 100.
Figure 5: Three route shortest path routing.
WEBIST 2019 - 15th International Conference on Web Information Systems and Technologies
270
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(a) document size = 10.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(b) document size = 50.
0
10
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90 100
processed ratio [%]
generate data speed [tag/s]
min10
cap10
acap10
lacap10
min50
cap50
acap50
lacap50
min100
cap100
acap100
lacap100
(c) document size = 100.
Figure 6: Two route disjointed shortest path routing.
5 CONCLUSIONS
In this paper, we have formulated and studied the
XML routing problem on a network that contains
XML processing capable nodes. We have defined
three single path routing strategies, and two types
of multiple path routing strategies, covering routing
without node repeats, and disjoint multiple paths.
We have evaluated these XML routing strategies
on an event driven simulated network, and verified
the following behaviors. On single path routing: i)
Minimum hop routing delivers poor network XML
processing performance; ii) considering available ca-
pacity prevents node overload condition; iii) intern-
ode latency may be used to spread XML traffic to
longer paths, delivering better network processing
performance. On multipath routing: i) Two route
path delivers better performance for large XML doc-
uments; ii) There is a saturation point, beyond which
an extra route does not cause performance improve-
ment; iii) Using disjoint paths does not improve XML
network processing performance, especially for link
costs which already take into account XML node traf-
fic loads.
Directions for future research include multiple
path XML routing for fault tolerance.
ACKNOWLEDGEMENTS
This work was supported by Japan and Technology
Agency (JST), Strategic International Collaborative
Research Program (SICORP), Japan.
REFERENCES
Cavendish, D. and Candan, K. S. (2008). Distributed xml
processing: Theory and applications. Journal of Par-
allel and Distributed Computing, 68(8):1054–1069.
Cavendish, D. and Gerla, M. (1998). Internet qos routing
using the bellman-ford algorithm. In Proc. of HPN
’98 Proceedings of the IFIP TC-6 Eigth International
Conference on High Performance Networking, pages
627–646.
Fan, Y. and Chen, A. (2012). Energy efficient schemes
for accuracy-guaranteed sensor data aggregation using
scalable counting. IEEE Transactions on Knowledge
and Data Engineering, 24(8):1463–1477.
Kalarani, S. and Uma, G. (2013). Improving the efficiency
of retrieved result through transparent proxy cache
server. In Proc. of fourth International Conference on
Computing, Communications and Networking Tech-
nologies (ICCCNT) 2013, page 1–8.
M. Shimamura, T. I. and Tsuru, M. (2010). Advanced re-
lay nodes for adaptive network services - concept and
prototype experiment. In Proc. of International Con-
ference on Broadband, Wireless Computing, Com-
munication and Applications (BWCCA) 2010, page
701–707, Los Alamitos, CA, USA.
S. H. Kim, K. Kim, C. L. and Ro, W. (2012). Offloading
of media transcoding for high-quality multimedia ser-
vices. Consumer Electronics, IEEE Transactions on,
58(2):691–699.
S. Nishimura, M. Shimamura, H. K. and Ikenaga, T. (2012).
Transparent caching scheme on advanced relay nodes
Efficient Shortest Path Routing Algorithms for Distributed XML Processing
271
for streaming services. In Proc. of International Con-
ference on Information Networking (ICOIN), page
404–409.
Solis, I. and Obraczka, K. (2006). In-network aggregation
trade-offs for data collection in wireless sensor net-
works. Int. J. Sen. Netw., 1(3/4):200–212.
Tennenhouse, D. L. and Wetherall, D. J. (2007). Towards
an active network architecture. SIGCOMM Comput.
Commun. Rev., 37(5):81–94.
Wang, Q. and Ozsu, T. (2007). An efficient eigenvalue-
based p2p xml routing framework. In Proc. of 2007
Seventh Intl. Conference on Peer-to-Peer Computing,
P2P ’07, page 105–112.
Y. Kanada, K. S. and Nakao, A. (2012). Network-
virtualization nodes that support mutually indepen-
dent development and evolution of node components.
In Proc. of IEEE International Conference on Com-
munication Systems (ICCS) 2012, page 363–367.
Y. Uratani, H. Koide, D. C. and Oie, Y. (2012). Distributed
xml processing over various topologies: Characteriz-
ing xml document processing efficiency. In Web Infor-
mation Systems and Technologies, volume 101 of Lec-
ture Notes in Business Information Processing, page
57–71. Springer Berlin Heidelberg.
Ziyaeva, C. E. G. and Min, D. (2008). Content-based in-
telligent routing and message processing in enterprise
service bus. In Proc. of 2008 Intl. Conference on Con-
vergence and Hybrid Information Technology, ICHIT
’08, page 245–249.
WEBIST 2019 - 15th International Conference on Web Information Systems and Technologies
272