This paper investigates a new algorithm for flocking of multi-agent systems with virtual leader. In this algorithm, unlike the existing algorithms, receiving of information by agents are not continuous and it is assumed that each agents receives information of neighbors and virtual leader as sampled data in fixed time intervals. A sufficient condition on sampling period is derived to guarantee the velocity convergence of agents to that of virtual leader, preserve network connectivity, and ensure collision avoidance. Finally, a numerical simulation is provided to illustrate the theoretical results.