Nowadays, people need to communicate with each other in each time and place. Therefore, cellular wireless networks have been developed to serve users. Considering the astonishing growth in the number of cellular networks’ subscribers, researchers are looking for the ways which can improve networks services. Recently, they proposed some new effective methods which can strongly satisfy user’s expectations. For instance, they introduced Long Term Evolution (LTE) for the third generation of mobile networks in order to provide the higher data rates for sending information in mobile networks. They believe that the newly introduced technology (LTE) can significantly improve frequency spectrum efficiency and reduce delay in the set up process. Also, the necessity of providing services with suitable quality is deeply felt. In other words, the subscribers expect to receive their services without any flaw. Also, providing services with the existence of failure in the networks has a signifant importance. For this purpose, different methods were introduced for the provision of resilience against network failures and the provision of network services in the case of failures. The majority of the previous methods suggested the increment in the Mean Time Between Failures (MTBF) and the decrement in the Mean Time To Repair (MTTR) as an effective way. In this research, a causality graph is used to show the causal relationships between special failures and services in LTE networks. Also, a special type of failure in LTE networks is mathematically modeled and the effect of this failure on the network performance is verified. The results show that it is possible to control the average waitin time with the arrangement of failure and repair rates. Also, two ways are suggested for the provision of fault-tolerance in cellular networks. One of these methods uses ad hoc relaying for helping users in the network at the moment of failure. In fact, this method suggests that it is useful to place a considerable number of Ad hoc Relay Stations (ARSs) in order to exchange the necessary signals between Mobile Stations (MSs) and Base Stations (). The simulation results show that the dropping rate of active connections is strongly improved by using the proposed solution, specialy when the number of ARSs in the network is unlimited. The other method uses bandwidth degradation in the adjacent cells of the failed one to reduce premature dropping rate of current calls. Also, it is possible to use the proposed bandwidth degradation solution in the bandwidth allocation process. The results show that the proposed method reduces the blocking and droppin rates in comparison with previous methods. Key Words: Fault-Tolerance, Cellular Networks, Bandwidth Degradation, Markov Model, Ad hoc Networks.