In grid networks, distributed resources are interconnected by wide area network to support compute and data-intensive applications, which require reliable and efficient transfer of gigabits (even terabits) of data. Different from best-effort traffic in Internet, bulk data transfer in grid requires bandwidth reservation as a fundamental service. Existing reservation schemes such as RSVP are designed for real-time traffic specified by reservation rate, transfer start time but with unknown lifetime. In comparison, bulk data transfer requests are defined in terms of volume and deadline, which provide more information, and allow more flexibility in reservation schemes, i.e., transfer start time can be flexibly chosen, and reservation for a single request can be divided into multiple intervals with different reservation rates. We define a flexible reservation framework using time-rate function algebra, and identify a series of practical reservation scheme families with increasing generality and potential performance, namely, FixTime-FixRate, FixTime-FlexRate, FlexTime-FlexRate, and Multi-Interval. Simple heuristics are used to select representative scheme from each family for performance comparison. Simulation results show that the increasing flexibility can potentially improve system performance, minimizing both blocking probability and mean flow time. We also discuss the distributed implementation of proposed framework.