In the most basic VPN scenario, a home user with a DSL modem, for example, can establish a VPN connection that forces all of his or her system's traffic through the VPN tunnel to a workplace network. This traffic includes everything from email and other corporate services to simple Web browsing.
When split tunneling is introduced into the equation, only a portion of the traffic is tunneled. Administrators configure the VPN tunnel to be network-aware, and the user's VPN client then makes intelligent routing decisions based upon each packet's destination address. If a packet is headed to a system on the workplace network, it gets routed through the VPN tunnel. If it's destined for an external site, it goes through the user's DSL gateway directly to the destination host.
The decision to use split tunneling depends upon your specific business needs. If your goal is to secure traffic between remote users and the workplace, it's fine to use split tunneling. If you do so, however, you'll need to educate your users and ensure that they know which traffic does and does not pass through the tunnel; you don't want to give employees a false sense of security.
Why wouldn't we want to avoid split tunneling altogether? When you don't use split tunneling, users can't access restricted resources on their local networks. Consider again the case of our home user. If that user has a privately addressed file server sitting on the home network, it won't be accessible without the use of split tunneling. Also, if the enterprise has a large number of users following this model, it may not want to bear the burden of processing large amounts of traffic bound for other networks.
This was first published in August 2007