Mobile Ad-hoc NETwork (MANET), which is a concept of mobilerncommunication without a fixed communication infrastructure, is one of therncurrent emerging technologies gaining much attention from both researchers andrnusers. Because there is no as such a fixed responsible device router, eachrncomputing node is to act as a router and should be willing to forward otherâ€™srnpackets.rnRouting in mobile computing is a difficult task as we have a very dynamicrnnetwork topology. Another problem is, the nodes in ad-hoc network are usuallyrnlaptops and personal digital assistants and are often very limited in resourcesrnsuch as CPU capacity, storage capacity, battery power and bandwidth. Thisrnmeans, any routing protocol running on a mobile node should try to minimizernrouting or control packets and also be reactive.rnA research work group established by IETF, has set a number of requirements tornbe met by an ad-hoc routing protocol. According to these requirements, arnnumber of routing protocols have been proposed broadly into two mainrncategories, proactive and reactive.rnIn this thesis, a detailed discussion of routing protocols from both proactive andrnreactive group is presented followed by simulation work using a discrete eventrnNetwork Simulator called NS-2. Because these groups of routing protocols failrnto satisfy all the requirements of MANET under all conditions such as highrnmobility, big network size and large number of traffic sources, a new scheme ofrnhybrid routing protocol is proposed. This new routing protocol, called ZRP, is tornxrntake advantages from both proactive and reactive group of protocols. Itrnalleviates scalability problem of other routing protocols by classifying thernnetwork into manageable sizes called zones.rnZRP combines two different routing schemes in one protocol. IntrAzone routingrnuses a proactive protocol to maintain up-to-date routing information to all nodesrnwithin its routing zone. By contrast intErzone route discovery is based on arnreactive route request and reply scheme. Simulation results of this thesis workrnshows that ZRP performs well under all conditions especially when we havernlarge network size.