Communications and Network, 2013, 5, 34-38
doi:10.4236/cn.2013.51B009 Published Online February 2013 (
A Communication Middleware with Unified Data
Transmission Interface
Yu Zhang, Yuanbing Zhang, Xiangyuan Bu, Chao Zhang
School of Information and Electronics, Beijing Institute of Technology, Beijing, China
Received 2012
In a Modern communication network, data exchange become a complex problem because there usually exists a various
data formats due to the diversity and complexity of communication devices. In this paper, a communication middleware
with unified data transmission interface is introduced, which acts as the abstract communication layer in the heteroge-
neous distributed private communication network. Devices within the network only need to interact with the middle-
ware, instead of directly communicate with each other by various protocols and interfaces, which can reduce the com-
plexity and diversity of the heterogeneous network. This article describes the structure and features of the middleware,
and analyses the use of “buffer pool” in message sending and receiving process. The applications of this middleware
can reduce the complexity of heterogeneous network interoperation, and improve communication efficiency.
Keywords: Communication Middleware; Buffer Pool; UDP; XML
1. Introduction
With explosive development of telecommunication tech-
nology, a modern communication network is becoming
more and more convenient and powerful. Meanwhile, it
becomes more and more complex, which brings addi-
tional technical difficulties. For a communication net-
work in a local area which integrates multi abilities, in-
cluding the telecommunication, localization and naviga-
tion, we must make the data transmission and exchange
between any two devices smoothly and reliably [1].
Since the network is composed of various devices with
different protocols and physical layer specifications, to
provide the direct connection between devices through
the physical layer becomes very complicated and costly.
Also, because of different data formats, the quick and
correct data format translation is crucial for the network.
To solve this problem, a communication middleware
with unified data transmission interface is introduced in
this article. Combined with the hardware adaptors, the
middleware acts as an abstract communication layer, and
provides an effective and flexible centralized message
service. It can receive messages from source devices, and
transfer them to the destinations accurately. As a result,
multiple devices within the network only interact with
the middleware through messages, and do not need con-
sidering the format converting or data translating process
[2]. The use of this middleware can greatly reduce the
complexity of device interoperation, and furthermore
improve communication efficiency.
2. Architecture of the Network
In this article, the communication middleware with uni-
fied data transmission interface makes the information
exchange between various devices of the network based
on a logical software bus, so the communication-related
devices only need to establish a direct connection with
the middleware, and can achieve communication with
other ones. Thus, the coupling degree between devices
within the network can be significantly reduced, and data
can be transmitted in a unified way. The architecture and
data flow of the entire communication network are
shown in Figure 1. System data are preserved in XML
format, and then transmitted through UDP protocol [3].
The information of various devices is saved in a system
database, so the middleware can obtain message con-
figuration information by querying the database, and then
transmit messages.
3. Basic Function of the Middleware
3.1. Centralized Message Service
*This work was supported in part by the Ministry of Science and Tech-
nology of the People’s Republic of China under Grant 2011BA-
It is the key service of the middleware. As a transmission
middleware of a large number of received messages, it
Copyright © 2013 SciRes. CN
cannot transmit them directly at the same time. Therefore,
the middleware creates a cache folder in the local server,
which is used to save th e received messages temporarily,
and transmit them from here as soon as immediately. The
establishment of cache folder is automatic and immediate,
that is, a message is automatically saved in the temporary
folder when sent to this middleware, and its forwarding
and query are both from here.
3.2. XML Format Messages
If the identifiable message formats of various devices are
not the same, a destination must translate data from
source terminals into a format that it can understan d, and
the translation process will cost a lot of system resource.
So the messages within network are all defined in XML
format, which simplifies the complexity of the transmis-
sion and exchange [4]. The XML file is represented by
some hierarchical elements. An element is defined by a
pair of tags, called the start and end tags. Content be-
tween the pair is the element body, which may contain a
set of child elements. The unified-format message is not
only easy to be transmitted, but also conveniently parsed
out according to certain rules when arriving at destina-
tion, and the message itself will not be changed.
3.3. Loose Coupling
This middleware is loosely coupled with other devices,
that is, it has no knowledge of other separate devices’
definitions. Devices and the middleware do not know
each other’s working process, and the communication
between them is completely achieved by the messages
[5]. As long as the message accords with the structure
regulation, the clients’ requests or the servers’ feedback
service can be realized. Also, should the receiver appli-
cation fail for any reason, the senders can continue unaf-
fected, as the messages they send will simply accumulate
in the message queue for later processing when the re-
ceiver restarts. Loosely coupling middleware platform
can make full use of existing system resource and con-
struct flexibly and effectively. Under the premise of no
more than the equipment load, the middleware can access
a large number of devices, in order to meet the require-
ment of large communication networks; it also can pro-
vide a powerful global informatio n organization al ability,
to reduce the error rate of the network.
3.4. Real-Time Transmission and Real-Time
This middleware uses the UDP protocol to transmit mes-
sages. UDP uses a simple transmission model without
implicit handshaking dialogues for providing reliability,
ordering, or data integrity, that is, a kind of connec-
tionless-oriented data transfer protocol [6]. Compared
with TCP, it has higher transmission efficiency, and is
more applicable to the real-time system. Also, UDP's
stateless nature is also useful for terminals answering
small queries from huge numbers of clients. And it sup-
ports packet broadcast (sending to all on local network)
and multicasting (send to all subscribers), which app ly to
the system with various message types and large trans-
mission capacity.
3.5. Persistent Messages Guarantee Reliable
As we know, UDP provides an unreliable service and
datagram may arrive out of order, or go missing without
notice. Therefore, both the sending devices and the mid-
dleware itself, use the programmable interval timer (PIT)
when sending messages. The timer’s interval is set as the
repeat sending time of message, so the same message can
be sent periodically after the timer starts, until the re-
ceivers return a finishing flag, then an interrupt is trig-
gered, and the next message can be transmitted. As a
result, a small amount of message loss will not affect the
information transmission.
4. Architecture of the Middleware
In this communication network, each device can be used
as a message sending or receiving terminal. These ter-
minals are able to widely listen for the location informa-
tion, their own state information, and the unexpected
information in emergency, and then transmit the infor-
mation to this middleware periodically. The middleware
will send them promptly, so as to ensu re that the destina-
tion terminals can receive updated information. The key
architecture of the middleware is shown in Figure 2.
Accurate and timely sending and receiving of mass
data is still the main problem the communication mid-
dleware faced with. To solve the problem, the technology
of “buffer pool” is used in this article, combined with
multicast to send and receive messages [7]. It contains
three key services as follows:
Figure 1. The achitcture and data flow of the communica-
tion network.
Copyright © 2013 SciRes. CN
4.1. Monitoring Control
This module completes the monitoring and receiving
functions. Ensuring to receive messages uninterruptedly,
the receiving port of middleware is always on. Messages
will be automatically received to the local server when
coming to the middleware. In the communication system,
the number of message is large while the sen ding time is
uncertain, so the middleware cannot send all messages at
the same time. Therefore, it is necessary to design a
message receiving buffer pool.
First, some memory space is opened up for receiving
buffer pool, and a receiving-operation-related list, Re-
ceive File Manager List, is set here, whose elements are
the message receiving threads. All the threads are defined
as the objects of the receiving class (Receive File Man-
age). The class relationship diagram of receiving buffer
pool is shown in Figure 3.
The Receive File Unit is defined as a message record.
The Receive File Manager is defined as a message re-
ceiving container class, which contains a dynamic array
object. The Receive File Manager List is defined as a
buffer pool class, that is, the message receiving buffer
pool. As is shown, the buffer pool class, Receive File
Manager List, contains 10 Receive File Manager’s ob-
jects, whose dynamic array object is the Receive File
Unit record.
Buffer pool is a kind of FIFO data structure. When a
new message arrives, the list will automatically check
whether the message exists in the buffer pool. If not, the
message receiving thread will be created and added into
the list; if it existed, new thread will be no long er created.
Starting receiving a message means adding a Receive
File Manager object into the buffer pool Receive File
Manager List. Only when the thread object is added, the
Figure 2. The key achitcture of the middleware.
Figure 3. The class relationship of the receiving buffer pool.
whole message receiving process begins, including the
file identification, unpacking, checking and resume bro-
ken transfer. The message can be completely received by
the middleware only when all of the operations men-
tioned-above finished correctly.
According to the time order, all the monitored message
thread objects will join the receiving buffer pool, which
completes the receiving operation one by one. And then
the threads will be released, so as not to take up too much
system resource. If the thread cannot be access to the
buffer pool timely and normally, the list will automati-
cally detect whether the buffer pool reach es its maximum
capacity. If so, it gives a blocking prompt to the message
senders. Using the buffer pool to manage the receptio n of
messages, no matter how large the volume is, can make
sure them received by the middleware platform orderly
and timely.
4.2. Message Configuration
This module configures the corresponding relationship
between the messages and destinations based on the
background database of the middleware. As the middle-
ware is connected with all communication devices in the
network, its background database has saved all their
connection information, such as IP address and port
number. When a message is received, this module will
configure its transmission ro uting acco rding to its tag and
information of the receiver. When the receiver estab-
lishes a connection with the middleware, this message
will be transmitted according to the configured routing.
4.3. Transmission Control
This module completes the message transmission func-
tion. Similar to the monitoring module, it uses connect-
ing buffer pool to manage the terminals. That is, a Udp-
Send List buffer pool is defined to manage all the Udp-
Send objects. The class relationship diagram of sending
buffer pool is shown in Figure 4.
Figure 4. The class relationship of the sending buffer pool.
Copyright © 2013 SciRes. CN
Copyright © 2013 SciRes. CN
Figure 5. The implementation of the middleware.
When the middleware is connecting with a terminal, it
will create a UdpSend object to package IP address, port
number and any other properties of UDP connection, and
add it to th e UdpSendList buffer pool. The b uffer pool is
conducive for the middleware to judge a terminal is on-
line or not. The terminal’s UdpSend object in the list
means it online, and the next sending operation can start.
When there are relatively fewer receiving terminals,
the UDP point-to-point transmission method can be used,
which means the platform operates the UdpSend object
of the UdpSendList directly. However, when sending
messages to a large number of receivers at the same time,
the multicast technology should be used. Before sending
a message, according to message routing configuration,
the middleware platform adds all the receivers that need
it into a multicast group. When th e middleware sends the
message to this group, all the receivers will get it. This
multicast group is dynamic, and the involved receivers
are changing depending on the message routing. The
middleware has a repeat sending time to ensure every
message can be sent to the corresponding receiving ter-
minals, and avoid the information omission because of
UDP packet loss.
5. Implementation of the Middleware
The implementation of the middleware is show in Figure
5. The implementation of the middleware contains the in-
terface driver control, routing configuration, traffic moni-
toring, user authority management and log recording.
Therefore, the middleware can connect devices within
the network through different interfaces and protocols,
and the messages from them can be configured according
to the routing and then received and forwarded. The
Figure 5 shows the log of data exchange, in which the
entire work process can be recorded. As is shown, a
sender keeps sending messages to this middleware,
which are processed by the internal buffer pool, and the
receiver can get them accurately and timely. The data
traffic is monitored by the middleware, in case of the
information congestion. Also, users signed in are man-
aged strictly. The middleware realizes the functions
above, and ensu res all of the data not delayed or lost.
6. Conclusions
A unified-data-interface communication middleware is
presented in this paper, which has a kind of centralized
message sending and receiving mechanism, and acts as
the abstract communication layer in a heterogeneous dis-
tributed network. Any two communication devices of the
network do not communicate with each other directly,
but through messages exchanged by the middleware,
then complete the communication functions such as
sending request and receiving service. Therefore, the
coupling degree of network is greatly reduced. Also, uni-
fied data format avoids the consumption of format con-
version, accelerates the flow of data, and thereby in-
creases the efficiency of communication process. With
the lower communication complexity, each device can
focus on their respective functions such as data process-
ing. For a communication network with fixed internal
communication requirements, large data traffic and
higher node mobility, the middleware platform is widely
[1] A. Tsutsui, H. Maeomiti, R. Kawamura and K. Yata, “An
Adaptive Communication Middleware for Network Ser-
vice Coordination,” In Proceedings of 1st IEEE Consumer
Communications and Network Conference, New York:
IEEE, 2004, pp. 406-411.
[2] J. AI-Jaroodi, N. Mohamed and J. Aziz, “Service Ori-
ented Middleware: Trends and Challenges,” In Proceed-
ings of ITNG2010-7th International Conference on Infor-
mation Techonogy: New Generations, IEEE Computer
Society, 2010, pp. 974-979.
[3] L. Caviglione, G. Ciaccio and V. Gianuzzi, “Architecture
of a Communication Midleware for VANET Applica-
tions,” In Proceedings of the 10th IFIP Annual Mediter-
ranean Ad Hoc Network Workshop, New York: IEEE,
2011, pp. 111-114.
[4] V. Schuermann, A. Buda and J. F. Wollert, “XML-based
Middleware Approach for Industral Wireless Communi-
cation Systems,” In Proceedings of 13th IEEE Interna-
tional Conference on Emerging Technologies and Fac-
tory Automation, NJ, USA: IEEE, 2008, pp. 632-639.
[5] G. Kitagata, J. Sekiba, T. Suganma, T. Kinoshita and N.
Shiratori, “Communication Mechanism of Loose Coupled
Agents in Fames,” In Proceedings of Seventh Iternational
Comference on Parallel and Distributed System, Los
Alamitos: IEEE Comput. Soc., 2000, pp. 467-472.
[6] G. Gehlen, F. Aijaz and B. Walke, “Mobile Web Service
Communication over UDP,” In Proceedings of IEEE 64th
Vehicular Technology Conferrence, VTC-2006 Fall,
IEEE, 2006, pp. 2876-2880.
[7] J.-Y. Chwng, D. Ferguson, G. Wang, C. Nikolaou and J.
Teng, “Goal-oriented Dynamic Buffer Pool Management
for Data Base Systems,” In Proceedings of First IEEE
International Conference on Engineering of Commplex
Computer Systems, Los Alamitos: IEEE Comput. Soc.
Press, 1995, pp. 191-198.
Copyright © 2013 SciRes. CN