Cloud data centers have a large number of resources. Management of such huge amount of resources for a large number of consumers requires fail-safe algorithms and leasing policies. Advance Reservation (AR) leasing policy is a rigid policy, which needs resource and consumer locking at a very early point of time, while advanced reserved lease can be rejected at actual point of time when resources are required. This problem can be dealt with proposed Improved Advance Reservation (IAR) algorithm and leasing policy , which uses negotiation and provide half capacity of the requested number of resources, instead of rejecting a lease if consumer agrees for the same. Experimental results show that the proposed work maximize resource utilization and acceptance of requests in comparison with existing algorithms in Haizea.