Method and apparatus for traffic optimization in virtual private networks (VPNs). A client device establishes a first VPN connection with a first server based on first VPN credentials. Traffic is transmitted and received through the first VPN connection to and from the first server. A second server is identified based on traffic optimization criteria that need to be satisfied by the VPN connection. Upon receipt of the identification of the second server the client device is to use the second server as a destination of a second VPN connection. The second VPN connection satisfies a set of traffic optimization goals for at least one flow from the flows forwarded through the first VPN connection. Based on the identification of the second server, the client device establishes the second VPN connection for the flow between the client device and the second server.