The invention discloses a query workload estimation-based extensible markup language (XML) fragmentation method, which comprises the following steps of: (1) coding each node in an XML tree by adopting an interval code Zhang coding rule; (2) generating a related XPath query step for each node, and adding the XPath query steps into an XPath queue; (3) recursively estimating the query workload of each node from the root node of the XML document tree by adopting a depth priority traversing sequence; (4) dividing the XML document tree into sub trees, the query workload of which is W0, according tothe query workload estimation result; and (5) sequencing the separated XML fragments according to the query workload estimation values, and distributing the XML fragments to the processing nodes in a'double-square' mode. By the method, the query workload estimation is performed only by using an XML document structure, query of a user does not need to be known, and the query workload estimation values are used as XML storage measurement, so that better query load balance and extensibility support are achieved.