The initial specifications of Web services cope with the issues of service publishing and service discovery but not with the issue of service selection. Service discovery, handled by UDDI and WSDL, alone is not sufficient to find the most appropriate server that can deliver customers' required quality of service (QoS). In this paper, we consider a broker-based approach to provide QoS support in Web services and deal with the selection issue. The broker may implement various selection policies that can range from static policies to dynamic ones, which take into account the current state of servers. Besides, servers may deliver different levels of service to their customers. We model this QoS broker-based scheme by a multi-class queuing model and we study its performance with a probabilistic splitting policy for server selection.