A lot of work has been done to evaluate the real time applications' performance over wireless networks. Some work focused on delay as a performance metric. However, no work studied the delay random behavior over a test-bed which imitates as much as possible actual operative networks. In this paper, we designed and run different experiments over heterogeneous 802.11 Wireless Mesh Network (WMN) test-bed with a scope of studying the delay probability distribution over multihop WMN. We anticipate that by knowing the link delay distribution, we will be able to use this information to design sound algorithms to provide for QoS. For example, designing a partitioning algorithm capable of partitioning multiple end-to-end QoS requirements into link QoS requirements. We designed different experiments to measure the link and end-to-end delays over the WMN test-bed. The measured link delays are used to construct an empirical histogram. Our experiments reveal that irrespective of the number of hops along the paths and type of traffic crossing the link, the empirical histograms almost have same general shapes. The empirical histograms are fitted into different types of standard pdf distributions. We found that the best fits for almost 90% of the empirical distributions are two standard distributions; gamma and logistic.