Grid Information Service Based On Mobile Agent

Grid is a kind of important information technology rised in recent years ininternational, which is a new-style network compute platform appeared afterWorld-Wide-Web. Grid uses Internet or appropriative network to make theextensive distributed, isomerous, and dynamic resources connect together, furthermore, it realizes highly sharing and integrating resources, providing high-poweredcompute, management, services functions and so on for users.Grid Information Service(GIS) is an important module to Grid. Gridintegrates all kinds of resources and realizes managing and accessing to theseresources efficiently by GIS. It actions as a bridge for information applicants andinformation providers in Grid. GIS needs to supply basic operations for acquiringinformation, offering information, and managing information. GIS provides somebasic functions, such as information registration, information updatement,information querying, information logout and so on.Currently, international researches on GIS are very active. There are forekinds of GIS: Globus GIS, relational GIS, GIS in Condor project, and GIS basedon P2P.Existing GIS can be abstracted and simplified as including three roles.Information server provides Grid resources information for organizing, storing,updating, querying, logouting function and so on. It builds a bridge for informationapplicants and resource providers. Resource nodes are Grid information providers.They can dynamically join in and quit out Grid organization. Users are applicantsfor Grid resource information. Information server needs to deal with the registerrequests from resource nodes, respond to query requests from users, even toanalyse and manage the dynamic information submitted by nodes periodically atthe same time. When the resource nodes number goes up quickly, informationserver will be the bottle neck of the whole system. In addition, resource nodeshave to run a dynamic information collection program to submit dynamicinformation periodically after their registrations. This adds work burdens toresource nodes.Aiming at the above problems, this paper brings forward Grid Information Service Based on Mobile Agent (MA-GIS). The primary thought is that makinguse of the mobile agent to fulfill the collection and management of information, inorder to lighten the work burdens of resource nodes and information server.This paper does the following detailed jobs:1. Using for reference the role sort thought of existing GIS, MA-GIS alsosorts Grid entities to three roles. Users (a general designation for the objects usingGrid information, concretely including the legal Grid users, Grid applications, andother function models in Grid system), Information server(collects and stores thestatic information and the dynamic information from the registered resource nodes,accepts the query requests from the users and returns the query results, signs thehost certificate as CA), Wide area resources(dispersive and isomerous entities atgeography location, including computers, clusters, computer pools, instruments,equipments, storage establishment and so on). MA-GIS organizes the three roles inthree layers hiberarchy, and sets up the general framework of MA-GIS.2. Designing the work process for MA-GIS under the above framework.(1) Resource nodes register to information server. It shows that they want tojoin in Grid environment. This registration includes two steps. Firstly, nodes applyfor host certificate, signed by information server(CA), and then submit registrationinformation to information server.(2) Information server stores the resource information to the informationdatabase. These information include node IP or hostname, information shared time,information shared rules and so on. Registered resources become to Grid resources.The information provided by them can be queried by legal users.(3) Information server collects information in two forms. One is thatinformation server periodically dispatches a mobile agent to each registered nodein order to complete the special information collection task, then perfects orupdates the data stored in information database, and waits for users's queries. Theother is that after the users forward their requests, information server dispatchesexpert mobile agent according to users's queries to collect information.Information server can dispatch created original agent to one resource node inorder to complete the task of information collection, also can dispatch multi-agentcopied from original agent to different or same nodes in order to completeinformation collection of different tasks, thereby realizes that collection tasks execute parallelly.(4) When users need resource information, they forward query requeststhrough browser. The interface provided for users must be simple and friendly. Itconceals the details of information collection. Users can only see the abstractedlogic resources. This make the query process vitrification.(5) Information server accepts the users query requests, then searches theresource information matching with the users requests in the information database,returns query results to users at last.(6) When resource nodes want to quit out Grid environment, they can throughlogout, request the information server to delete their information records from theinformation database. When resource node quits out Grid unconventionally,mobile agent can not arrive to that node, it can record this condition in its interiorstates, and returns to information server to delete the resource information. Thisensures the validity of the resources information.3. This paper uses IBM Aglet as the development platform of mobile agent.The transfer mechanism is that holding the transfer route and transfer condition ofmobile agent hidden in the task code of agent. And it adopts less spending butmore efficient weak transfer. MA-GIS adopts itinerary pattern to realize theinformation collection of mobile agent. Information server periodically dispatchesmobile agent taking message to each resource node, and these messages aremanaged at the resource node.4. The information that MA-GIS collects mostly hidden in kernel informationof the operating system, such as the length of running process, idle memoryamount, CPU speed, machine load and so on. This paper mostly considers how tocollect the information of the machine running Unix and Linux os.5. In the LAN environment of lab, author uses six PC to build and simulateone Grid environment, does a simple simulation experiment for MA-GIS.MA-GIS does one improvement to GIS. It is that GIS doesn't need resourcenodes to collect dynamic information, also doesn't need information server toanalyze and dispose information, but makes the mobile agent dispatched byinformation server complete these tasks. It not only lightens the work load ofresource nodes and information server, but also unifies the information collectionbecause the collection policy is established by information server. In addition, mobile agent only needs to transfer a small quantity of codes and states to resourcenodes in order to collect and analyze the information, and return the results toinformation server. This avoids transferring vast resource information, accordinglyreduces network loads and saves bandwidth for other Grid activities. Whenresource node quits out Grid unconventionally, mobile agent can not arrive to thatnode, it can record this condition in its interior states, and returns to informationserver to delete the resource information. This ensures the validity of the resourcesinformation. When Grid organization members increase continually, MA-GIS willshow obvious advantages. So, applying mobile agent to GIS and making use of theadvantages of mobile agent technic to efficiently implement the GIS functions,accordingly provides information for other Grid activities, has a deep meaning inthe GIS researches.