This thesis proposes power control algorithms for maximizing the transmission rates in digital subscriber line systems where crosstalk is the major performance-limiting factor. First, a joint multiuser detection and optimal spectrum balancing algorithm is proposed for the case when the DMT blocks are perfectly aligned. Assuming that the set of tones subject to multiuser detection can be arbitrarily selected, this algorithm uses the idea of dual decomposition to jointly determine the optimal set of tones and the optimal transmit power allocation scheme. Second, we investigate the effect of DMT block misalignment. Inter-carrier interference is induced as a result. Heuristic algorithms are proposed to maximize transmission rates in this case. The main idea of the algorithms is to perform a neighborhood gradient search on the solution derived from an existing spectrum balancing algorithm. To reduce the algorithmic complexity, an approximation method using the first-order power series expansion is suggested.