VMT Design And Implementation-a Mobile Distributed System Visual Model Constructor

With the advancing step of the Internet, the environment of computer softwareis encountering significant changes from static and closed into dynamic and open.Here among the changes,"dynamic" means that the situation changes frequently, and"Open" refers to that the software environment and its changes may be further thananyone has ever expected. Since the hardware speed keeps increasing, the storagecapacity is extending, its cost is going down and the Ad-Hoc network and mobilecomputing devices comes into being, the scale and complexity of the softwaresystem have been greatly increased. Hence software system is migrating from thetraditional desktop environment to a new level with higher system distribution andmobile features.This paper firstly describes the background of mobile distributed systemcharacteristics and the importance of mobile middle-ware. Comparing with fixednetwork, the wireless network is more susceptible to interference and be easier toencounter network failures which would cause system disconnection and thendecrease the availability of the system. So it is quite necessary to adopt a supportingmethod for disconnections which includes: caching technology, data collection, RPC,deployment and re-deployment, copy consistency and code consistency. Among allabove, the deployment operation is to deploy the components to hardware devices.Moreover the middle-ware method is also enrolled to setup the distributeddeveloping platform which can provide support to disconnection operations.Middleware is able to check the running information of the system, hence amiddle-ware adapter component was designed to fulfill the requirements from themobile distributed network.Secondly, a mobile distributed visual constructor MTV was built by MVCdesign model. VMT has the following features: tailorability, scalability and effectiveness. The tailorability of VMT makes it possible to add a new view ormodify existing views. VMT also support scalability which takes place in the modelsize, view, and controller algorithm scalability. VMT’s effectiveness is to ensure thatthere is no cost in performance in exchange for system scalability. MVC modeldivides an application into model, view and controller. VMT model sub-systemincludes system data components, Algorithm Results data components and graphicalview data component; VMT view sub-system includes the table view componentsand graphics view component; VMT controller sub-system includes a servercomponent, server modifier components, algorithm container components andmiddle-ware adapter components. The model, view, and controller are all designedby J2EE architecture.Thirdly, the definitions of availability, model assumptions, availabilitycalculating method and deployment algorithm to increase availability were presented.Model Assumptions include: Although there are a lot of reasons that can causeavailability reducing, we focused on the availability of the system and the operationsof disconnections and connections, since the connection operations are main causesfor system performance loss. In order to increase system availability in a given timeT, we assume that the access to some intermediate position can be supported by thesystem deployment structure, i.e. some system-related parameters can be acquiredthrough some intermediate location. Availability model building constraints:limitations with available memory on each host, memory requirement for eachcomponents and component locations (for example, UI components may be a fixeddeployment to a specific host). Calculate the overall system availability under allabove constraints.Finally, because of the inner mobility and the unpredictable nature of the mobiledistributed environment, the hardware hosts and/or components may get intounforeseeable failures. If the hardware host failed, the components deployed on itwould stop their services or require service from other hardware hosts that wouldobviously reduce the system availability. If the system consists of n components andm hardware hosts, it is a question of index complexity to find the maximum systemavailability, hence an approximate algorithm is expected to implement thedeployment of the system. The pseudo codes of random deployment and greedydeploy algorithm were evaluated here. In the experiments, system availability and running time were collected and analyzed based on two different algorithms on3instances. Although greedy deployment algorithm has a little more time consumedthan the random deploy algorithm, it has a significantly higher system availability.