With the demand for more capacity, driven by the end user and the competition among network operators, it is inevitable that fibre-to-the-home (FTTH) deployments will become reality for most customers in the future. However, FTTH represents expensive solution for delivering broadband speeds due to the expensive labour, scattered population, problems with home fiber wiring, and slow roll-outs. As an FTTH alternative, hybrid fiber-copper access is proposed where fiber is brought as close as possible to housing complexes and then existing copper cables are used to bridge the last 20 to 250 meters which represents the most expensive part for fiber deployment. Legacy copper based technologies, i.e., digital subscriber line (DSL), are capable of delivering up to 100Mbps over 17MHz spectrum and long loop lengths (up to 1 km). This turned out not to offer sufficient capacity for broadband services. ^In order to address this issue, International Telecommunication Union (ITU) has initiated standardization of the new (fourth) DSL technology, called G.fast, which allows network operators bring speed of up to 1 Gbps over short copper wires (< 250 m). Crosstalk cancellation (also known as vectoring) represents an essential tool for G.fast systems to achieve the foreseen bit-rates. In this thesis I carry out a performance study of crosstalk cancellation algorithms (both linear and non-linear) in a G.fast-compliant setup by extensive simulations. ITU has agreed that for the lower-bandwidth profile (up to 100 MHz) linear vectoring schemes shall be used whereas there is an ongoing discussion whether linear or non-linear vectoring schemes shall be used for the 212 MHz profile. In my performance analysis I focus on the achievable bit-rates of both linear and non-linear vectoring schemes for 212 MHz profile. Furthermore, the considered vectoring schemes depend on accurate channel knowledge. ^Thus the influence of channel estimation errors on their performance is also investigated showing that even with 99.9 % accurate channel estimation (i.e., only 0.1 % estimation error) bit-rate losses are considerably high. Vectoring requires joint signal processing among the users at either the receiver or the transmitter side. However, in many deployment scenarios joint signal processing is feasible among sub-groups of users only, referred to as partial vectoring. Since vectoring cancels only crosstalk originating within one group, for previous DSL systems, such as Very high bit rate digital subscriber line 2 (VDSL2), dynamic spectrum management level 1 (DSM-L1) and level 2 (DSM-L2) schemes have been deployed to mitigate crosstalk originating among different groups. In this work I evaluate DSM-L2 for the case of partial vectoring in a G.fast-compliant simulation setup. ^I also propose interference alignment (IA) as an alternative to DSM-L2 and show that IA outperforms DSM-L2 when vectoring is not applied most notably on spectrum above 100 MHz. However, simulations indicate that in case of partial vectoring for G.fast simple time division multiple access (TDMA) outperforms both DSM-L2 and IA. Deployment of currently deployed DSL technologies with G.fast modems will be gradual and therefore the coexistence of G.fast and legacy DSL systems needs to be investigated before bringing the G.fast technology to the market. Due to the different modulation parameters (e.g., carrier spacing and sampling rate) joint vectoring between these two systems is not practical, raising the question in which setup G.fast and legacy DSL may coexist in the same binder. Therefore, an accurate performance model is crucial for assessing the impact of VDSL2 on G.fast and vice versa. ^Differences in modulation parameters along with (possibly) asynchronous transmission lead to inter-carrier and inter-symbol interference (ICSI). Thus, an ICSI model which captures all these differences is required for performance evaluations of mixed xDSL scenarios.Therefore, in this thesis I propose and investigate an analytical ICSI model for asynchronous discrete multi- tone (DMT) systems with different carrier spacing and sampling rate. I also combine developed ICSI model with CSI error model in order to obtain a joint performance model for mixed vectored DSL systems including the impact of ICSI and channel estimation errors. A detailed simulation study is carried out on the impact of ICSI on G.fast and VDSL2 systems. These show that bit-rate losses in G.fast and VDSL2 systems can be substantial compared to the interference-free performance. ^Further, the influence of different VDSL2 transmit and receive filters is analysed focusing on the required spectral separation between these two systems. The simulation results indicate that in order to have negligible interference between G.fast and VDSL2 systems, G.fast should start at frequencies not below 23 MHz assuming the high order filtering at VDSL2 transceivers.