Building optimizied web service based on service discovery technique and artificial algorithms ACO and HMM
Research
BUILDING OPTIMIZIED WEB SERVICE
BASED ON SERVICE DISCOVERY TECHNIQUE
AND ARTIFICIAL ALGORITHMS ACO AND HMM
Nguyen Thanh Long1*, Nguyen Duc Thuy2, Pham Huy Hoang3*
Abstract: Service discovery is important network function that has just developed
inherited from auto detected device function of Windows OS. Though the application
of this technique in service based routing is not trivial. Moreover, nowadays
computer technology has evolved, computer networks have grown to a new stage.
This network is called the upper layer network; the network nodes are
interconnected on the basis of connection according to the service. This paper
focuses on designing an application of service discovery on the basis of demand
analysis, object-oriented service oriented routing. With the evolving variety of Web
services today, it is essential to take advantages of these services to perform new
complex tasks. Based on the researches of some of the world's researchers by the
well-known articles [2], the author of this paper proposes a technique that uses
graph theory combined with ACO optimization algorithm, to create the optimal
route to support the service discovery problem. Based on the study of optimal
algorithms, the thesis inherited this research, simulating this research by optimizing
the optimization of the service on the basis of membership services, adding the
optimal algorithm Gen to create optimal routes and avoid optimizing locally.
Keywords. Service discovery service, Web service, Multicast routing, ACO, HMM.
1. INTRODUCTION
1.1. Some concepts and definitions
Explore the service information to find out, select the most specific information
of the services are set up on the network. These findings will be gathered at the
central control system or distributed across multiple servers across the network. On
the basis of this information, will create new services meet the complex
requirements of the user. From the diverse requirements in fact, the thesis found
that the need to form automated services through membership services, on the
basis of gathering the requirements from the user side. All newly created services
will be cached in the routers, so that they can be used as needed. Service
Information Discovery Mechanism: i) explore connectivity to shared software
systems; ii) Implement on the Internet through public port address; iii) Use
standard protocols and software to communicate and exchange information quickly
and reliably.
1.2. The development history of service discovery
Service-oriented routing is required to optimize service discovery to form a
database framework/platform to manage and exploit efficient network services.
Currently, service delivery systems are organized in the form of clustering and
decentralization. As the system of resources of the cloud, all organized the type
of service. Whereas the basic part is looking for services, in order for this
59
Journal of Military Science and Technology, Special Issue, No.57A, 11 - 2018
Electronics and Automation
search to work, we must build search tree types. These trees are managed
centrally or dispersed.
By the distribution model, the login information will be displayed on the
multiple node/transaction points and will be synchronized through the routing
protocol of layer below. The such register will need effective routing functions.
Now this multicast routing will be activated to report the service search request to
the service provider. The multicast routing is running in the upper network layer.
Follow the paper [3], for shrinking cost, energy consumed for service searching,
the paper mentioned the virtual network architecture for this service discovery, this
architecture also matches the upper class network architecture that has defined in
the thesis. The virtual network is an existing network files are a role of the core /
frame. This virtual network setup the network back-bone. This network frame also
makes the service discovery directories.
The author has suggested the concept of cataloging, local cataloging, and
cataloging of service descriptions. The service discovery request will be sent to the
local directory. If the requested service is not found, the local directory sends this
request with a choice to other categories in the global discovery process. The
engine of choice for this directory service is optimized on the basis of exchange of
records. This is a summary of the catalog and description of the capabilities of the
service server. These records are stored in standard formats such as XML for easy
exchange, archiving, and searching. For example, SQL Server also has strong
support for these data formats.
The author has brought to the concept of hybrid MANET network, which provides
adaptive search engines with boundary routing devices. When determining the
trajectory of moving nodes, the article [3] proposes to distribute service / service
registration messages in the orbit of each node. The nodes near the trajectory will
receive the message, which is usually the boundary router of each cluster. Actually
it is difficult to determine the trajectory of motion, which is based solely on
coordinates of nodes at a time around the current time using GPS.
1.3. Mathematical model for service discovery
F(Optimal Service Discovery)({Sꢀ, Sꢁ, …, Sꢂ}) = (Sꢃ ꢅ Sꢃ ꢅ … Sꢃ ꢅꢇ) (1)
ꢄ
ꢄ
ꢆ ꢆ
ꢇ
This is a formula that describes the results of a service search, assuming we
have services, which are stored on service servers, with service descriptions on the
categories. Service search results are a combination of member functions from the
service. The result of this search engine will be the combination of features of the
service that suits the requirements of the subscriber / device required. If each
service is considered as one axis of the q-space service, then the search space will
have the complexity of:
ꢊ
ꢉꢋꢀ
∏
O(
| |
ꢈꢉ )
(2)
Without using the search tree as mentioned in the introduction of the thesis, the
complexity will grow as network services become more and more diversified.
After mapping this domain to a graph, using the traditional algorithms on the
60
N. T. Long, N. D. Thuy, P. H. Hoang, “Building optimized web service ... ACO and HMM.”
Research
graph, the complexity is: O[N*Log(N)], N is the number of vertices in the graph.
We must use combinatorial optimization algorithms to optimize search engine
discovery services. Particularly artificial intelligence algorithms need to be studied
and applied. One of the easy-to-apply algorithms that are highly effective is the
Genetic Algorithm. Moreover, if we can map the hypothesis / condition of service
discovery to the problem of finding the optimal route Optimize on graph. On the
basis of graphs, traditional algorithms as well as artificial intelligence can be
applied to optimize search results. One of the most well-known artificial
intelligence algorithms for optimal routing is the optimal clustering (ACO).
1.4. Set up a service discovery system
Service discovery for high efficiency, article author [3] provides the principle of
constructing efficient lists, minimum energy consumption as follows:
Within the coverage of any node, minimize the overlap of the categories.
Catalogs periodically promote some status and summary data of service data
stored on the catalog.
After a certain time period if the button does not receive a service ad bulletin
for any category. The button will trigger the search process for you.
The thesis proposed to encrypt data exchange between the list and the
network node to avoid information exposure. The article [9] uses bloom
filters to exchange service advertisement data between catalogs and nodes,
making data transmission secure, efficient, and reducing bandwidth
requirements.
2. SERVICE FORMULATION ALGORITHM
As analyzed in the review chapter, the optimization service is mapped to the
optimal linear optimization problem on the ACO/Termite graph. ACO solves
online routing problems for telecommunication networks [10]. ACO is an active
routing algorithm. We can distinguish between: i) Routing in traditional IP
networks: optimal or shortest paths to the associated link/distance vector routing;
ii) ACO routing is adaptive. The article has confirmed that dynamic problems such
as routing in the telecommunication network using the principle of ACO algorithm
to solve is appropriate. The ACO's operational principle is to receive information
about routes through repeated sampling using special format packets to gather
information about latency and number of routes on each route. Therefore, it is
necessary to find the way to map the conditions of the problem to the fastest graph.
Here is the algorithm:
Proposed model for building optimal service from member services:
Constructing a directed graph G = (V, E) to simulate a model of services with
connected functions through constraints, each of which is characterized by a
function of a service. With this approach, the proposed architecture is easily
implemented in a programming language to demonstrate the feasibility of the
model. Each vertex of the graph v ∈ V represents a function of a certain service.
Each function of the service is characterized by two types of textures:
61
Journal of Military Science and Technology, Special Issue, No.57A, 11 - 2018
Electronics and Automation
i) Input structure:
This structure is formed by input parameter information, the number of input
parameters, the data type of each parameter, the meaning of each parameter;
ii) Output structure:
This structure is formed by the output parameter information, the information is
also managed as the input parameters above. The two vertices are linked together
when the result of the source vertex function is the input to the function
representing the vertex that is connected. In order for two functions of two services
to be connected, the condition is that the output structure of the derived function
must conform to the input structure of the link's function.
Thus, the problem of formulating the optimal service to satisfy the user's dynamic
demand is Y = F (X), where X is the set of input parameters, Y is the set of
parameters / outputs / rendered. F is a composite of service functions:
(X) = [ ꢌꢀ ∗ ꢌꢁ ∗ ꢌꢍ … ∗ ꢌꢅ ] (X) = Y
(3)
Consider the nature of the problem to find the solution to the function equation
(2) on a weighted graph, which requires the formation of an optimized route on
graph G. If this problem solves in the usual way is difficult NP problem. However,
in terms of the proposed thesis, the solution is only complex as evaluated for the
Greedy, Kruskal, ACO/Termite algorithms approximately: O [N * Log (N)], where
N is the complexity spatial data set of service information. If the route search
results are recorded logically, combining with one artificial intelligence algorithm
such as GEN for optimization is appropriate, the response time for optimal route
search is not large enough to match the Ad-Hoc network.
So the way to do this is to map the service find problem to find the optimal path
on the graph above. Using ACO/Termite to assign the probability of existence to
vertices related to routes is the solution of the problem. Thus, the most probable
route is the optimal route on this graph. Mapping the automated service formulation
problem to the optimal route finding problem on a network with directional weighted
graphs that are formed by ACO/Termite. Assess cost of links by the amount of time
required to execute function belonged to start vertex. Based on that to determine the
pheromone per link/peak that the ACO/Termite algorithm uses, thereby determining
the probability of choosing links from each vertex of the graph. Using ACO helps
reduce operating costs during optimal route search.
To execute the parsing model on, information about the service must be managed
centrally and distributed. If according to the distributed models, some hosts do the
same job to choose the optimal route. A generic manager server will collect results
and search best route to inform the client has given request. Each service includes
the functions or procedures executing its private tasks. This information will be done
by an administrator or automated by computer based on service description
information. When finding right information to send client and cache this result for
following searching to get the result fast. With any new service information or adjust
from provider the cache content have been adjusted or deleted out if no fit.
62
N. T. Long, N. D. Thuy, P. H. Hoang, “Building optimized web service ... ACO and HMM.”
Research
Search for routes is based on building multi tree, with root is the node that fits the
request. The leaf nodes will be the functions that make the matching results.
3. APPLY ACO TO OPTIMIZE SEARCH ENGINE
Based on the multi-path routing solutions mentioned, the problem will be to find
multiple multicast trees, with the root of each tree being the same topology structure.
Find the optimal route on the multicast trees
To find the optimal route, there are many methods, such as OSPF using the
Dijkstra algorithm. MANET networks have unpredictable network structure, so the
ability to use traditional algorithms is very difficult, and using these algorithms
requires a lot of computing. Nowadays, artificial intelligence algorithms have
brought about high economic efficiency. These algorithms use the actual situation of
the data to optimize. Although local maximum may occur, the execution speed is
only a linear function of the input data. Thus, for the highly adaptive routing method,
the paper proposes the application of ACOs to the transmitting multipurpose trees.
The tree propagation algorithm was investigated in the article [1], this paper presents
this algorithm in the context of the problem posed.
3.1. Graphical building algorithm from member services
Graphs need to be constructed as a directed graph. It is easy to see that the graph
can be clustered by the number of input or output parameters of the corresponding
function of each vertex. In addition, the list of input/output parameters of each
cluster can be aggregated into two ordered arrays (Lst_of_Input, Lst_of_Output) to
easily find the vertex that is connected to the considered vertex. Suppose we need to
add a vertex that is represented by function F to the graph:
i) Consider the set of output parameters of the function for this vertex:
ꢈꢎ = {ꢏꢀ, ꢏꢁ, …, ꢏꢐ}
(4)
Find cluster with each vertex that has parameter number matching number of
parameters in ꢈꢎ: Lst_of_Input. All elements of the array: Lst_of_Input, each item
with parameters are also ordered in ascending order. Then combining ordered
parameters of each item between each parameter is a separator (sc). Combined
parameters in ꢈꢎ have been sorted, we have:
ꢑꢒ = ꢏꢓꢀ[sc]ꢏꢓꢁ[sc]…ꢏꢓꢐ
(5)
If it is found that there is a link from the current vertex to a vertex in the cluster.
ii) Find input parameters of the function representing the by the similar above
method by the list of output parameters in the Lst_of_Output of a cluster. If it is
found that there is a link from a vertex of the cluster to this vertex.
For each element in an array with a pointer to a vertex, one can find the vertices
of the graph that are connected to the vertex just as easily. The complexity of this
algorithm is just: O [Log (N)], N is the number of vertices in each cluster.
Using the multi-path route finding algorithm as described in the article [1], with
the condition that the departure node of the route has given the input parameters, the
end of the route has the necessary outputs.
63
Journal of Military Science and Technology, Special Issue, No.57A, 11 - 2018
Electronics and Automation
In addition, after finding the desired trees, the ACO algorithm for the trees found
can be used to evaluate the probability paths. Get the most probable route to perform
the service. Assume that these routes have an existence probability changing over
time, particularly when the nodes of the route are not clearly defined. It is then
possible to use the hidden Markov model to manage these routes efficiently and
when it is necessary to predict the nodes that route pass through when the node
information is not clear.
3.2. Use ACO algorithm to rating the current traffic
Begin from the root tree, estimate the probability to choose the next vertex that
route passes through based on the 2 factors is: ꢔꢕꢖ(t): pheromone trail parameter and
ηꢕꢖ. In which, ηꢕꢖ was given by formula: ηꢕꢖ = ꢀ , with T is the current function j
ꢕꢖ
ꢗ
ꢘꢙ
function of i service. Therefore, the probability to choose next vertex depends on
pheromone accumulated by time and the time to perform function in normal
condition. The probability to choose next relay node is defined by formula:
ꢟ
ꢡ
ꢛꢜ (ꢝ)ꢞ ꢛꢠ ꢞ
ꢘꢙ
ꢘꢙ
pꢚꢕꢖ =∑
,
ꢡ
(6)
ꢟ
[
] [
]
ꢠ
ꢘꢢ
ꢜ (ꢝ)
ꢘꢢ
ꢇ
ꢘ
ꢢ∈ꢣ
In which, Nꢕꢚ is the set of vertices that each vertex has a link to the current vertex
but at which is not visited. At the beginning time, the root of tree is current node, so
the probability to accept in route is 1. With k is the symbol of the Ant is being
implemented.
The algorithm to find optimized route (on-demand web service) starts algorithm
from the source node, the algorithm chooses next nodes that have probability not less
than a threshold value to go. The algorithm scans tree until reaching a fitted node on
each branch of the found multicast tree or it couldn’t go further. The algorithm has
| |
complexity of O ( ꢤ ). V is the vertex set of the graph. However, when we need to
find a single optimal route, we will choose the maximum probability path. In
addition, we can divide the input data into subsets and perform on all paths found,
leading to the following goals: i) parallel execution; ii) make the most of the system
resources; iii) Lead to load balancing. The probability of existence of route is
calculated by the product of the probability of all the nodes that belong to the route.
According to the principle of ACO, the probability of selecting a node Nꢕꢥꢀ with a
previous vertex on the route Nꢕ is computed by the quotient of the pheromone of
node Nꢕꢥꢀ divided by the total pheromone of nodes that have a link to Nꢕ:
ꢜ
ꢣ
ꢘꢦꢄ
P(Nꢕꢥꢀ) = ∑
(7)
ꢧ
ꢜ
ꢣ
ꢙꢨꢄ
ꢙ
In which: (Nꢕ, Nꢖ) E.
The process of updating the pheromone is made after finding the routes that
satisfy the problem after a number of predetermined times.
ꢃ
(t+1) = ρτꢩ(t) + ꢚꢋꢀ ∆τꢚꢩ(t)
(8)
∑
64
N. T. Long, N. D. Thuy, P. H. Hoang, “Building optimized web service ... ACO and HMM.”
Research
Where: τ (t + 1) is the accumulated pheromone of the node at time t + 1, ρ∈ [0,
1]. This parameter is called the stability of the route, ∆τꢚꢩ(t) is the increment of
pheromone at time t of node N, m is the number of routes to which node N
belongs. If the node doesn’t belong to any route, pheromone is updated according
to the formula:
(t+1) = ρτꢩ(t)
(9)
From the above formula we can see that the pheromone of non-routed nodes will
progress to 0. The next section of the thesis presents the hidden Markov model to
improve service quality for the problem of finding the path. The ACO mapping to
HMM, the HMM implementation using graphs, the node location prediction model
uses HMM to apply in clustering and find paths.
4. FIND THE MULTIPATH ROUTE BASED ON
THE HIDDEN MARKOV MODELS
4.1. Research problem
The hidden Markov model is a mathematical model, apply statistical
probability to simulate the operation of any physical process. The evaluation, find
route in the MANET network will be easily optimized when using HMM to
simulate. Because the MANET network has links that are setup and change very
fast, so that it is often difficult to implement and not highly adaptable. Based on
this analysis, the thesis proposes the use of HMM to predict some information in
the network with fast-changing configurations over time. Most process mobilize in
fact can be modeled or represented by hidden Markov models. Since the
connections in the MANET network may not be clearly defined by the influence of
the environment, the use of HMM to estimate the probability of route existence in
some particular cases is very necessary.
Propose the uses of HMM to reach some goals:
- Predict the position of the network node on the basis of article reference [4]:
According to [4], each node, after receiving a neighbor finding message, will
store this node in its neighbors table and calculate the probability of the node
(fuzzy logic) within its coverage area based on: i) distance between two nodes; ii)
energy remaining; iii) signal strength. The nodes in the coverage area with the
probability of reaching the new threshold can forward the packet.
- Predict the link status as shown in the article [5][6].
Related studies:
Effective routing in a mobile network with intermittent connections:
In paper [7], the authors propose optimal routing solutions for network with
non-persistent links. This is the case with networks that have frequent
disconnections that may be due to energy savings in wireless sensor networks,
possibly due to oscillation around the equilibrium position. For the high probability
of receiving packets, the source node needs to transmit each message a number of
times. Therefore, the number of messages received at the destination node can be
redundant, but it will prevent the message from being lost due to a lost connection
65
Journal of Military Science and Technology, Special Issue, No.57A, 11 - 2018
Electronics and Automation
at the right time of transmitting the message from one source node. Assuming that
at the source node, all links are ineffective, retransmission of the message is
mandatory. Just as the above analysis can use HMM to predict the location of
nodes in this network it is appropriate to find the route of data transfer as needed
and to complement the existing routes as they lose connectivity.
Multicast routing in a structured overlay (P2P) network:
Multicast technology is very popular in advanced networks, including P2P
networks that use this technique to improve performance. Multicast is a service
that is managed and optimized by service-oriented routing. In the upper layer of
the network to ensure the quality of service is the most important, it should use the
routing in coordinates to improve the adaptability of the existing routing algorithm
is appropriate.
Markov model of links in the MANET network:
In paper [8], the author proposes a state transition model of nodes, where
conditions are met, the node switches state. Transferring this state will set up or
break the network links. It is necessary to build the state file of each node, the state
transition of the node. Then guess the probability of linking at each node. As with
HMM for node prediction, the use of HMM to find the probability of link existence
is also of concern in such networks as MANET.
4.2. Describe the working principle of HMM on the basis of probability
graphs
HMM works on the basis of some state sets, among which there is a state
transition probability:
There is a set of derived states: this is the state set whose existence exists at
the beginning of a defined probability (S).
This is a set of states that the current process is going through (H).
The set of end states is the states that contain the result of the problem (F).
Upon departure, each initial state has a probability of existence that the process
may belong to. On the basis of the network parameters, the state transition will
occur to give an end state which is the result of the problem. The thesis
recommends using Fuzzy Logic to evaluate the probability of this existence. In
addition, we can map the states of HMM to the vertices (V) of the graph (G), the
ability to shift between states mapped to the probability of existence of the (E)
linkages of G. Therefore, we can construct a directed state transition graph for
HMM, based on this graph to improve the performance of the HMM. Based on the
minimum tree search algorithms, colorize, to find the optimal search tree and
transmission tree for anything in the real world. In addition, we can use ACO to
update the probability of linking between peaks over time.
4.3. Perform ACO problem using HMM
From the definition of HMM, it is possible to map the optimization problem on
the graph using ACO as the maximum probability problem on the HMM to move
from start state (ꢈꢎ) to end state (ꢈꢪ). Managed networks are structured as a
directed graph with each vertex having links with the probability of existence
66
N. T. Long, N. D. Thuy, P. H. Hoang, “Building optimized web service ... ACO and HMM.”
Research
defined by ACO/Termite mapped to the HMM model as analyzed in the preceding
section.
Map the set of vertices of the network graph to the states of the HMM
system under review. Map each vertex of the ACO as a subgraph in the
HMM. These are the states in which the node belongs.
The probability of the state transition between any two states I and J in HMM is
the probability of linking from node I to node J in the ACO graph (corresponding
to two nodes of two subgraphs in HMM). as determined by ACO/Termite.
Routing tables defined by ACO/Termite are also mapped to the state transition
matrix (A) of the HMM, but each node can belong to a set of states in the HMM.
So finding the optimal line between any two vertices ꢫꢬ, ꢫ in ACO/Termite
ꢭ
is mapped into the problem finding the maximum probability of moving the
state from initial state I to target state J.
Where: I, J are the states of the observed L states, the intermediate vertices in the
line are the hidden states of the set N. One example, in the HMM graph, optimized
from the source node to the set of target nodes, at this point the source node and
target node set are the states of L.
4.4. State the problem
Design a hidden Markov model to predict the location of network nodes to
reduce control costs by reducing the transmission of control packets when
rebuilding routing table supports optimal route detection in mobile ad hoc
networks.
The meaning of the solution:
In case there is no routing table or the remaining energy is small, we must
predict the location of the node for transmission based on the neighboring table
(NT).
Input: A network graph consisting of nodes, links between nodes is
determined based on the state of the nodes received from the NT table.
Output: Find the multi-lane optimization route in a fast-moving network such
as the MANET network.
Algorithm: Use the hidden Markov model [4] to predict the position of the
node based on the collected state of the nodes. Uses HMM to predict the mobility
of each network node. This is done at a base station or cluster node in each cluster
in a network of small nodes such as sensors or clustering networks. If the network
is large and unconnected, this is done at each node. Information about the mobility
of each node is predicted, then sent back to the nodes in the cluster and stored in
the coordinate table (CT). When a node wants to transmit data to another node: i)
The source node takes the coordinates of the nodes in the cluster of the CT table;
ii) If the destination node belongs to a cluster, the node transmits data directly to
that node in coordinates; iii) If the node does not belong to the cluster, then the
data is transferred to the cluster head by coordinates. The cluster head will base on
neighbor cluster information to propagate the message as well as the coordinates.
67
Journal of Military Science and Technology, Special Issue, No.57A, 11 - 2018
Electronics and Automation
4.5. Uses HMM to predict the position of the network node
To find the route between the source and destination node, the source node
must propagate the RREQ route packet. RREQ is only transmitted to the node with
a probability of existence greater than a threshold value in the coverage of the
source node under the HMM model. The process continues like this with multiple
paths to the destination node. Initially the source node transmits the message
requesting neighboring nodes to inform it of its future location [4]. Based on that
node will determine the probability of each neighbor node in its coverage area to
decide whether to send the incoming RREQ message. Repeat process until
destination node, route is formed. Initially, each node can belong to one of the N
states, with each state having a probability P_i. During operation, each node can
move state to a set of unobserved states with probabilities that can change over
time. These states are predictable based on the information received from the node,
such as signal strength, trajectory, time elapsed from the beginning. For example,
when we know the signal strength we can determine the relative distance to node
(d), so the node will be located on a sphere of radius d. It is easy to see that the
position of the node is the orbital intersection of the node with the sphere. In this
problem the state of the node is indicated by the coordinates of the node. Each
node can move to a set of observed states, here are some visible positions of the
predefined node based on observation or node information. If you do not receive
the information from the node, you can predict the current position based on the
previous position of the node and the node parameters. The parameters received
from the node are used to deduce where the node can belong to which state with
the highest probability. The parameters received from the network node are the
results in HMM (F). One example might be that the parameters received from the
network node are the signal strength.
The simulation showed that the routing results using HMM gave better results
than DSR, with the number of hop lines decreasing.
4.6. Use HMM to find the path in the clustered network
For clustered networks we can look at the whole network as a hidden Markov
model. Where each cluster is a state of the HMM, at which point each node can
belong to a cluster with a probability. This probability can be determined by the
relative position of the node relative to the cluster head in question. It is also
possible to use the coordinates of a node as defined above to allow any node of a
node. Initially, the node belongs to a particular cluster of the network with a
probable probability. Based on the coordinates of the predicted node, when no
cluster node information is received, we can predict based on the previous state of
the node and the received node information such as the velocity of travel of the
node compared to other cluster heads. It is easy to find paths in the probability-
based network, at each cluster if the destination node is not in cluster, select cluster
with cluster head that has a total distance to the current node and the shortest
destination to transmit the packet.
68
N. T. Long, N. D. Thuy, P. H. Hoang, “Building optimized web service ... ACO and HMM.”
Research
5. SSI
to findd the broad
ious seerv ce orientation.
functioon and rocedur
dently before pper orming the tre
m to evaluate thee optimal route sele
time p riod, or ere the route
ill study additioon l optim zation
s of adaptive rouuting in conjunction with artific
ulation s writtenn i C# pro rammi g languag
of real Web sser ices, so simul tion results can
ion and businesss at the workk unit o the author.. The si
service setup was conduuct d a number of ti es, measuring the time takken to graph
the res lts. Finding the liin s satisfyiing the roblem fromm the gr ph is forrm d, usin
the AC algori hm to fiin the pro ability of each ro te bein selectedd. rom th
find pa hs with probability choose the best path that as the h ghest prob bility o
creatin a Web service.. As a resullt of the simulation progra , the timme taken t
perfor all t_All jobs annd he opti al rout search, the generated Web sse vice wa
created with the requestt b ing pas ed thro gh the t_EEx interf ce to thee f le. Logs
and gr phs sho the timme variability of t e algorithm in so e impleem ntations
plemenntat on is reported re interfacce:
M
M
ULATION A
ast multiples menti
raphs ar
NND ASSESSMENT
Use
t
he grap
h
cc
o
e
o
ned in
t
he previio
n the bbas
s that cca
. Usingg t
t is perffor
. In thee f
e optimmal route o
gence allg rithms.
, the input of thee s stem is
e used to recover.
ulation of the We
u
u
s sectio
s of th
be ru
e AC
ed ove
ture, th
nn
ee
nn
OO
rr
ee
and in the pre
structu e of th
indepe
algorit
v
e
i
i
G
built
oo
i
i
r
n
hh
s
s
p
ees of Web servic
e
mm
n
n
ff
e
search algorith
h
h
cction probability th
a
m
m
a give
author
the bas
n
n
e
whh
m
aa
m
eets certain criteri
lgorithms to find t
al intell
a
a
uu
w
w
aa
i
h
h
nn
ii
i
ii
oo
yy
The sim
i
n
n
g
nn
e
aa
numbe
rr
vv
a
a
bb
produc
tt
ff
mm
bb
ee
mm
uu
ee
pp
aa
ee
gg
O
t
dd
b
u
g
FF
ee
t
li
,,
hh
i
ro
aa
ff
g
mm
oo
mm
tt
m
e
rr
ss
e
e
ss
uu
aa
i
i
a
w
hh
mm
e
e
..
Each ti
m
m
e the im
i
i
s
s
ults on the programm
z
y
y
x
Figur
e
e
1. Comparison of Y-axis exxecution time (graphh gener
service implemeent tion) wi h time ( ind optim l route,
axis) self-service motion.
a
i
a
tion, opt
mplemenntation) (X--
tima
i al route
search
,,
a
a
t
ff
aa
Lo
o
o
king at the graphh a
b
b
ove, wee find th optimal rooute and
e
e
s
s
ervice pper ormanc
f
f
ee
is quite stable o
v
er time an
d
d
the ratio of time ꢝ
e
ꢮꢢꢢ varies in a rel
aa
tively staable range.
e
ꢝ
ꢯꢰ
699
Journal
oo
f Military Science aan
d
d
Technol
o
gy, Speci
a
a
l Issue, No.557A, 11 - 22018
Electronics and Automation
6. CONCLUSIONS
Service discovery is an important function in the upper layer network, so that
this function works optimally, especially in order to reduce energy consumption
and network control costs. The thesis proposed a solution for intelligent network
segmentation, multi-path routing, and multiplexing, to enable this function to be
effective. Map the problem finds efficient, optimal optimization problem to find
multi-line route, optimization on the graph, to ease the use of search trees, reduce
costs. On the other hand, this method makes it easy to use the ACO algorithm to
optimize and evaluate the probability of existence.
REFERENCES
[1]. Nguyen Thanh Long, Nguyen Duc Thuy, Pham Huy Hoang, “Building
Multiple Multicast Trees with Guarranteed QOS for Service Based Routing
Using Artificial Algorithms”. Springer, ISBN: 978-3-319-29235-9, ICCASA
2015, LNICST 165, pp. 354–369, 2016.
[2]. Sami Zhioua (2013): “Tor traffic analysis using Hidden Markov Models”,
Security Comm. Networks 2013; 6:1075–1086.
[3]. B.G.Obulla Reddy, Maligela Ussenaiah, Rayalaseema (2012): “An Adaptive
Fuzzy Clustering and Location Management inMobile Ad Hoc Networks”.
International Journal of Computer Applications & Information Technology
Vol. I, Issue III (ISSN: 2278-7720).
[4]. R. Nagwani, D. S. Tomar (2012): “Mobility Prediction based Routing in
Mobile Adhoc Network using Hidden Markov Model”. International Journal of
Computer Applications (0975 – 8887), Vol. 59– No.1.
[5]. Seok K. Hwang, Dongsoo S. Kim (2007): “Markov model of link connectivity
in mobile ad hoc networks”. Telecommun Syst (2007) 34:51–58.
[6]. Amardeep Sathyanarayana, Pınar Boyraz, John H.L. Hansen (2008): “Driver
Behavior Analysis and Route Recognition by Hidden Markov Models”.
Proceedings of the 2008 IEEE International Conference on Vehicular
Electronics and Safety, Columbus, OH, USA.
[7]. R.Sharma, D. K. Lobiyal (2015): “Proficiency Analysis of AODV, DSR and
TORA Ad-hoc Routing Protocols for Energy Holes Problem in Wireless
Sensor Networks”. 3rd Inter. Conference on Recent Trends in Computing.
[8]. Jaspreet Kaur, Dinesh Kumar, Giani Zial Singh (2017): “Distributed Hash
Table based Routing for P2P Data sharing in MANETs”, e-ISSN: 2454-6615,
WWJMRD 2017; 3(7): 82-85.
[9]. Franc¸oise Sailhan, Val´ erie Issarny (2009): “Scalable Service Discovery for
MANET”. HAL Id: inria-00414946.
[10]. Gianni Di Caro (2004): “Ant Colony Optimization and its Application to
Adaptive Routing in Telecommunication Networks”. PROMOTEUR: PROF.
MARCO DORIGO, ANNEE ACAD´ EMIQUE´ 2003-2004.
70
N. T. Long, N. D. Thuy, P. H. Hoang, “Building optimized web service ... ACO and HMM.”
Research
TÓM TẮT
XÂY DỰNG MÔ HÌNH DỊCH VỤ WEB TỐI ƯU TRÊN CƠ SỞ KỸ THUẬT
KHÁM PHÁ DỊCH VỤ VÀ CÁC THUẬT TOÁN TRÍ TUỆ NHÂN TẠO
Khám phá dịch vụ là một chức năng mạng mới phát triển từ chức năng khám phá
thiết bị trong hệ điều hành đã có trước đây. Nhưng việc áp dụng hiệu quả kỹ thuật
này trong định tuyến hướng dịch vụ là một việc không tầm thường. Hơn thế nữa
ngày nay, công nghệ điện toán phát triển, mạng máy tính đã phát triển đến một giai
đoạn mới. Thế hệ mạng này gọi là mạng lớp trên, các nút mạng liên kết với nhau
trên cơ sở kết nối theo dịch vụ. Trong bài báo này, chúng tôi đề xuất một kỹ thuật sử
dụng lý thuyết đồ thị kết hợp với thuật toán tối ưu ACO, để tạo ra tuyến tối ưu hỗ trợ
bài toán khám phá dịch vụ. Trên cơ sở nghiên cứu các thuật toán tối ưu, chúng tôi
đã kế thừa nghiên cứu này, mô phỏng nghiên cứu này bằng bài toán hình thành dịch
vụ tối ưu trên cơ sở các dịch vụ thành viên, bổ sung thêm thuật toán tối ưu Gen để
tạo ra các tuyến tối ưu nhanh và tránh được tối ưu cục bộ.
Từ khóa: Service discovery service, Web service, Multicast routing, ACO, HMM.
Received 27th June 2018
Revised 10th October2018
Accepted 27th October 2018
Author affiliations:
1Trung tâm Công nghệ thông tin, Viễn Thông Hà Nội;
2Viện kỹ thuật Bưu điện, Học Viện Bưu chính Viễn thông;
3Viện Công nghệ thông tin, Đại học Bách khoa Hà Nội.
71
Journal of Military Science and Technology, Special Issue, No.57A, 11 - 2018
13 trang
12/04/2022
5440
Bạn đang xem tài liệu "Building optimizied web service based on service discovery technique and artificial algorithms ACO and HMM", để tải tài liệu gốc về máy hãy click vào nút Download ở trên
File đính kèm:
- building_optimizied_web_service_based_on_service_discovery_t.pdf
