Transaction scheduling has long been a crucial issue in database system design. Concurrency control algorithms used can have significant impact on system throughput. Past studies indicated that both pure blocking-based and abort-based protocols were unsatisfactory. Blocking harms a system by holding system resources whereas transaction aborts lengthen system response time because of restart workloads. A delicate balance is required to attain optimal system performance. To date, there is still no in depth study on the way that blocking and abort affect the performance of transaction scheduling for database systems. In this dissertation, a critical evaluation of the effects of the balance between blocking and abort in database system has been carried out for both centralized and distributed database systems with special reference to the problem of multigranularity lock interference and client distribution. In a centralized system, multigranularity locking allows each transaction to use the granule sizes most appropriate to its mode of operation so as to reduce the cost of locking. However, the result is the higher probability of lock conflicts as transactions with coarser granularity can be in conflict with transactions using finer granularity of lock. The degradation of system performance due to this lock interference cannot be underestimated. In the Study, the effects of balancing the degree of transaction blocking and abort on multigranularity lock interference have been investigated. Another area of investigation is in the effects of the balance between blocking and abort on the communication overhead for distributed database systems. Because of the increasing lock holding time due to internode communications, commitment protocols and possible delays in deadlock resolution, the balance between transaction blocking and abort is more critical to distributed than centralized systems. In connection, two important parameters communication delay and client distribution have been examined. For both centralized and distributed database systems, the Study has found that the cost of transaction scheduling can be significantly reduced by a careful choice of the balance between blocking and abort. The Study also shows that lock conflicts resulting from multigranularity lock interference can be attenuated through a better balance between blocking and abort.