Optimized Architecture-centric Function Points’ Clustering and Aggregating About Service Flow

This study puts forward the “Cluster” concept about service function points. For the uniform of timely information’s capturing, logical optimization and requirement implementation, service topology’s relationship is adaptively formed. Tracking architecture goal and building distributed clustering logic timely, “Cluster head” and “Cluster node” are serialized and optimal architecture-centric “Cluster group” is formed, it is to achieve the balance service adjustment of service topology network driven by goal.


INTRODUCTION
Control the complexity of software business and improves software services' quality, in order to raise the layer of software reuse (Mei and Shen, 2006).The specificity of software architecture in software life cycle and key role depend on the medium between users' requirement and goal's implementation (Garlan, 2000;Xiao-Feng et al., 2010), on architecture design, it is to achieve a flexible platform from a group of requirements to satisfy requirement or helping to meet for requirement by design (Bosch, 2000).
Architecture-structured and object-oriented technology provides a strong support for design, but it is difficult to adapt to the self-learning requirement of platform business logic and the high-layer special question of architecture.Architecture-structured stresses the core effect of function, while it does not achieve the uniform of question space and answer space and it is difficult to adapt to requirement changes, it expresses the "bloated" and "detailed" top-down requirement, but does not apply to high-layer design of architecture (Shao, 1998;Parnas and Clements, 1986).Requirement and business logic is planned by object, which unifies the modeling method between question and answer, but cannot be suitable for expressing the uncertain relationship among objects and granularities' selforganization, it exists serious "shortcomings ripple effect" for the more adaptive requirement, especially the non-functional features and integrated system properties are not be reflected in the architecture (Sangwan et al., 2008).
Architecture-centric service-oriented mechanism: In the architecture-centric platform, services is no longer the calculated granularities that is passively waiting goal requester's searching and adjustment, but can be real-time adjusted and take the initiative to search for architecture goal's implementation.In this process, it builds the corresponding relationship between goal requirement and service entities and then provides service function.If the goal is satisfied for service entity, they can collaborate on requirement.the goal's providing and solving is the topology of services' capturing and implementing sequence of architecture, then accepts the topology rules' logic and achieves the process and service mechanism about architecture goal.From the architecture perspective, it is the relationship of goal and services, from platform perspective; it is the aggregation of service topology and optimal building about goal, as shown in Fig. 1.
Figure 1 is divided into three parts, the middle is the intermediary platform (ACP: Architecture-centric Platform), it is to achieve the service process of Platform Requirement (PR), in the whole process, whole implementation logic is that architecture goal (AG) submits and feedbacks results, which exists parallel relationship between PR and AG.All business logic is the integration of service granularities (AG), that needs the aspects' supporting: first, AG of ACP achieves requirement logic's transferring of PR sets and completes logic's separation and aggregation, then forms the basic underlying topology relationship factors.Meanwhile, about service Topology Relationship sets (TR), Topology Relationship Analysis mechanism (TRA) analyzes these factors and selects the least and most simple relationships, cross reference, to complete the basic relationship factors' selection about PR implementation logic; Second, driven by basic relationship factors, helped by AG, these factors enter into Service Container (SC), Active Service (AS) captures the double deployment of goal's intention and Service flow's definition: Architecture achieves the transferring process from tasks' requirement to implementation sequences that is the process mode about requirement of platform, here gives the definition of service flow: Service is the basic granularity about implementation sequence and results' responses, controlled by architecture-centric main goal; this process is defined as service flow.We can see from Fig. 1, started from platform requirement, service flow accepts adjustment of architecture goal and finishes the corresponding and composition relationship about services and flow topology.In entire process, architecture goal plays an important role, which the topology corresponding and granularity's composition is an adaptive learning and completing transparency sequence.Service sets and architecture goal's achievement needs SC's integration and perception.Meanwhile, it is the basic supporting of service flow and the optimal implementation result.Every service flow's implementation is about goal's completing, this process is also responsible for adaptive services' selection and composition, the improvement and expansion of basic flow's topology.
This study puts forward the "Cluster" concept about service function points.For the uniform of timely information's capturing, logical optimization and requirement implementation, service topology's relationship is adaptively formed.Tracking architecture goal and building distributed clustering logic timely, "Cluster head" and "Cluster node" are serialized and optimal architecture-centric "Cluster group" is formed, it is to achieve the balance service adjustment of service topology network driven by goal.

SERVICE'S CLUSTERING LOGIC
The architecture-centric software platform based on services need to solve two question: function points' aggregation and service topology routing's building, its goal is to enhance the accuracy of service selection and the optimal degree of routing, at the same time, improve the ability that services participate in platform requirement's goal and achieve the entire adjustment of AG about whole and local requirement, which has the same similarities with WSNs about platform's design and implementation.Inspired by the clustering of WSNs, this study proposes the service cluster concept.
Cluster: Cluster structure is divided into two kinds of basic units: Cluster Head and Cluster Member.Cluster head is the adjustment center of cluster structure and manages cluster's forming; data's collection of cluster member and collects the gotten data and transfers to other cluster heads or directly sends to dada aggregation point.
During the process of forming clusters, through corresponding algorithm, the neighboring nodes' information of cluster head is exchanged and judges the intensity of nodes' distribution, based on this values, the covering area of nodes in cluster is calculated, by adjusting all cost by broadcasting information, it makes the dense area form clusters with small area, oppositely, form cluster with large area.Meanwhile, the selection of cluster head is based on the residual energy of nodes, when selecting cluster head, the direction of cluster head and the residual energy of cluster head is as the precondition and chooses the best cluster nodes to join.
This process to achieve WSNs topology based on clustering is beneficial to the optimization and nodes' selection for whole architecture and routing.In architecture-centric service platform, service is as the basic participation granularity, adjusted by AG, the adaptive transferring process is achieved that expresses "business goal-optimal topology-service node".Here presents the service cluster concept that AG is statute and optimal service topology is the clue.Based on AG, architecture goal is divided into local ones, then service cluster head, cluster node and cluster group is defined formally.
Service Cluster Head (SCH): Service cluster head is the service individual about the Sub Goal (SG) for AG by cluster head's selection is to participate in the cluster nodes of SG, then SG is completed by service clusters.Service cluster head transfers the implementation results, or communicates requirement and results about SG with other service clusters.The service cluster head of architecture is formally defined as: , of which, i stands for the node number of first-class service cluster head, j stands for node number of second-class cluster head and so on, the final class n stands for the service cluster head number of AG.The achievement of AG is the topology's optimization and implementation of service cluster head set is defined as:

, ,
Service Cluster Node (SCN): Service cluster node is the participation node of cluster based on SG, that is the individuals and selected and adjusted by SCH, then takes part in implementation and achievement of SG.It is formally defined as SCH< i, j>, of which, i stands for cluster head node's number corresponding to common service cluster nodes, j stands for number of common nodes.For entire AG, service cluster head and cluster node exist such a relationship: ) ( is the cluster concept about the function point set, cluster is the structure unit about service cluster head and node, its physical structure is the individual service, cluster is the aggregation of SG, SCH and SCN, it is formally defined as : SCG< SG, SCH, SCH>.

GOAL-DRIVEN CLUSTER PARTITIONING STRATEGY
The forming process driven by achieves hierarchical cluster in a recursive way, using QoS value of every service cluster node, is divided into set of, ensure the cluster heads of different clusters, according to the association coupling of cluster heads, the participation nodes and boundaries are formed.
• The service granularity uses nodes to spread around their QoS characterization and service features regularly, builds corresponding service function list.
Service is adjusted by and provides its function and the collaboration information, so that, every service granularity will get candidate granularity group, through QoS and collaboration requirement information, the local has been calculated.• About cluster structure with aggregation relationship logic, analyzes QoS characterization and service features of every node, ensures cluster head nodes and gives corresponding to cluster head.• Based on the logical function of goal, analyzes every cluster head node, according to value, the association of local clusters and interconnection relationship among cluster heads are built.• The first-class cluster about is built: the cluster head nodes achieved by the above three steps is broadcast clustering information in architecture and form their own clusters.After service nodes' receiving broadcasting cluster information, the matching about local cluster and is verified.If it has joined some cluster, the virtual interconnection about functions are built and the logical association is built among clusters.If it has not joined, this node directly join cluster and gets the participation identifier.Of which, expresses cluster head node's number, expresses the number in the platform.• And so on, achieves the cluster and the association topology rules in or among clusters, when new requirement goals are provided, firstly search whether there're precedent instances in historical topology rules and improves the implementation efficiency of platform.If existing parallel service requirement goals, executes the first four steps.The final implementation topology structure is shown in Fig. 2.

OPTIMAL AGGREGATION PROCESS OF SERVICE CLUSTER
Starting from service quality of user requirement, with meeting for AG, service granularities with less response time and cost and larger success rate, reliability and composite rate will be selected for clustering.Therefore, when processing new user requirement, the optimal aggregation and the first-class cluster's forming, that needs the five QoS performance indicators to be ordered and selected by right, their sorting and selection is adjusted by AG.
Service cluster is the set of basic granularities, that meets for production rules and interface semantic rules of service composition model, according to these rules, the optimal aggregation of cluster processes corresponding the next step action a, inputting I and outputting O to form definition 1 a C , furthermore, based on the interface semantic rules of service composition, the participation interface semantic rules of cluster is aggregated to The optimal aggregation of service cluster is the balance and statute for second part of this study, here, suppose the QoS performance of service clusters is the definition above, respectively: While, ( all  Return (C g )//C g is the optimal aggregation result about service clusters.End

TOPOLOGY'S BUILDING OF SERVICE NETWORK
The optimal aggregation result of service clusters is to meet for the user requirement's implementation topology of AG, the cluster head nodes need some transmission power to transfer goal to all service clusters and nodes.After service nodes receive corresponding goal and requirement, AG calculates implementation routing distance based on QoS of selected cluster and nodes and forms service topology network of AG.
The forming of service implementation is the multi-layer distributed cluster routing with balancing QoS value, in order not to increase the difficulty about topology's forming and goal's implementing and improve the efficiency of routing's building, service cluster's forming is based on the optimal aggregation strategy of the fifth part of this study .Because the cluster head nodes directly receive the adjustment of AG or SG, it plays a key role on cluster's selection and routing's matching, during keeping to optimal aggregation strategy, this study will think about QoS performance of cluster head nodes.Therefore, routing topology needs the clustering of WSNs and routing topology algorithm, based on clustering and clustering's selection thought, the first-class service cluster head nodes are as the carrier about main goal In multi-layer topology of Fig. 3, service nodes in the same ring adjacent to each other, every ring is firstclass service cluster structure, the black nodes in middle part is the forthlayer service cluster node set, these cluster heads is the sub-class cluster head services about different SG, subclass cluster heads can be divided into one sub service cluster structure by goal, the small clusters in Fig. 3 is division of sub-class cluster head.During the forming of routing, new cluster heads will be produced from current non-cluster head nodes of implementation topology.That is, the next layer's cluster heads are not in the first-class service cluster head set, the service cluster head in this current layer only processes the logical aggregation and integration of this layer's SG.
Through different levels' service cluster heads, participation service nodes assign and select its own neighbor service node list.During the forming of the hierarchical topology, with the help of AG, from the whole to the local or from top to bottom, the cluster head nodes are selected.That selects first-level cluster heads and forms the first-level cluster, through traction, until forms k-level and this level is as the basic single service nodes, this completes multi-layer routing topology.By Algorithm 2 that can achieves the multilayer topology, in the selected and aggregating process

Fig. 1 :
Fig. 1: Architecture-centric service serialization accepts services' Classification Mechanism (CM) and AG, finishes the requirement's integration among services.The whole relationship is the transferring and analysis of AG, ultimately, it is the adaptive relationship's matching and implementing of services.

Fig. 3 :
Fig. 3: Multi-layer tolology of service routing and topology and build requirement-driven platform's implementation sequence.To reflect the adaptive capturing and timely feedback, multi-layer second-class service cluster head nodes are built to implement local goal SG.In multi-layer topology of Fig.3, service nodes in the same ring adjacent to each other, every ring is firstclass service cluster structure, the black nodes in middle part

−
At the same time, every service cluster is as the service aggregation structure, it has this five QoS performance values, that is defined as:Output the optimal aggregation of service clusters and the optimal association among service clusters, supposing the goal service composition structure of user requirement is C g .