I cannot agree with the suggestion that ASSM is "slower for high-volume concurrent insertion". Completely the reverse. The algorithm for searching the bitmaps (which is based on hashing each session's SID) means that concurrent searches for suitable blocks will tend to search different parts of the bitmap, which will point to different blocks of the segment. So no buffer busy waits on the bitmaps or the data blocks. The old freelist mechanism is awful in comparison, with sessions queueing up for access to free lists and the data blocks at the high water mark. In a RAC, it is even better because the instance number is included in the algorithm, so you tend to get an affinity of parts of the bitmaps to each instance.