Policies & Procedures

Port Splitters

Generally, all network entities (computer, networked printer, or other electronic "network aware" device) must be connected directly to a "network port" (i.e., a wall port or a wireless access port). Central network administration will assign that entity an appropriate IP address.

Whenever in-building network facilities are upgraded, both building users and network designers should take care to install sufficient connectivity to allow for both current needs and reasonable expansion, so that port-splitters need not be used after the upgrade.

Discussion & Procedures

The number of network connection points demanded by users has increased dramatically since the original computer network wiring was installed in UM-M buildings. Adding additional connection points by connecting a port-splitter (AKA mini-hub) to an existing connection point provides what seems to be a low cost alternative to the rewiring required to add new connection points, but adds some specific hidden costs.

• First, for heavily loaded networks the effective bandwidth for devices attached through a port-splitter is rougly that of the original port divided by the number of "split out ports." (In fact, the effective bandwidth may be even less because of inherent inefficiencies that increase as the number of devices connected to a single port increases.) As users demand higher and higher effective bandwidth for network applications (e.g., involving audio and/or video), splitting the available bandwidth with even one other device can cause performance problems.
• Second, central monitoring and troubleshooting is significantly more complicated when multiple devices with different IP addresses share the same physical wall port.
• Third, connection of multiple devices to a single network segment (e.g., through a port-splitter to a single network port) allows a knowledgeable user to monitor all of the traffic passing over that segment, so it is inherently insecure.

Thus, though it may be initially cheaper to attach devices through port-splitters, the result is a network that costs more to maintain, delivers lower performance, and may pose substantial security risks.

It is clearly not feasible to simply "outlaw" the use of port-splitters, given that there are many buildings and work-group setups within buildings where current wiring supports fewer connection points than are demanded by users. However, as the University moves forward with in-building network upgrades, particularly those involving use of wireless access as a cost-effective alternative to rewiring, the use of port-splitters should be avoided unless there is no other alternative. In those cases, explicit written approval should be obtained from the IT Office.