Abstract: The technique of dynamic interval index is very important in the practice of constraint database. Based on the structure of Meta tree, DM tree is given. Its storage is O（n） and query I/O time is O （log n+t/B ）. To reduce the complex of insert and delete operation, methods of unfixed constraint and late update are applied. The complex of the insert and delete I/O time are all O （log n ）. This result is equal to that of B tree. It can be used in the storage and query in constraint database.