In order to evaluate the characteristics of fire smoke dispersion in large-space buildings, FDS software was used to investigate the effects of air-inlet position, smoke-proofing depth and smoke exhausting rate to the working efficiency of fire isolation belt. The results show that, by opening the air-inlet in the non-fire zone and closing that in the fire zone, a reverse airflow can be induced below the fire isolation belt, which resists the smoke overflow and reduces the disturbances to the smoke layer. The increase of the smoke-proofing depth can increase the space for smoke storage and reduce the height of the smoke layer, thus delaying the instant of smoke overflow by decreasing the generation rate of fire smoke. In the scenario without mechanical smoke exhausting in the fire isolation belt, the instant of smoke overflow can be delayed by about 50% when the smoke-proofing depth increases from 20% to 40%. By adopting mechanical smoke exhausting in the fire isolation belt, the accumulated smoke can be timely extracted, which can effectively prevent the smoke spilling into other non-fire zone. It is concluded that, through an overall consideration to the smoke isolation efficiency, building functions and construction costs, the combination of "40% smoke-proofing depth + adopting the mechanical smoke exhausting rate with reference to the Smoke Standard + opening the air-inlet in the adjacent non-fire zone" is recommended as an advantageous design principle for fire isolation belts.