In IP network, the coarse packet classification is a scalable QoS solution. Similarly, it can also be utilized to OBS network because of its scalability. However, the OBS network can support only few priorities of optical burst due to the limitation of current optical device and signal process technology. Hence, in an edge node of OBS networks, the bursts are usually assembled through the many-to-one composite class burst (CCB) assembly technique, e.g., N:1-CCB. Currently, the QoS issue in OBS is as active as that in IP networks. The burst assembly has a great impact on the QoS support of OBS network. According to the study of the N:1-CCB burst assembly technique, there are two aspects related with the scalable QoS support, i.e., the mapping relationship and the assembly resource allocation. This paper simultaneously focuses the two aspects, and based on the linear optimization theory, proposes a novel assembly mapping mechanism, i.e. Optimized TQ-MAP, which most important feature is adaptivity. Based on nonlinear programming and differential calculus, it allocates the burst assembling capacity between classes fairly, efficiently and differentially, and matches IP QoS requirement with OBS QoS capacity as possible as it can. The simulation results show that Optimized TQ-MAP is more adaptive, and can efficiently guarantee the coherence of QoS support from IP to OBS.