In this paper, we consider the impact of traffic burstiness on optimal batching policy for energy-efficient Videoon-Demand (NVoD) services. By introducing batching technology, multiple users can be served by one multicast transmission. The more users in one multicast transmission, the less transmissions are needed, which induces less channel occupancy and energy consumption. However, to expect more users coming leads to longer queueing time. Hence, there is a trade-off between the number of transmissions and the average queueing time of users. We prove that N-Policy is optimal among all feasible policies when the arrivals of users follow a Poisson process. When the inter-arrival time of users follows Gamma distribution, i.e., the arrival process is smooth, it is shown that N-Policy is still optimal. But, when the inter-arrival time of users follows hyperexponential distribution, i.e., the arrival process is bursty, the N-policy is no more optimal and the energy cost can be further reduced by a Generalized Impatient (GI) policy, especially when the coefficient of variance is large.