In the IT world, a Single Point of Failure (SPoF) is a single vulnerability of the infrastructure. A SPoF can be both of a hardware (switches, routers, servers, single server components, etc..) and of a software nature, and its role is so important that it could compromise the whole system in case of failure. To draw an example, a local network that is connected with the outside world with a single, not redundant router has a SPoF in that router, as a failure or the breakage of such router completely stops any LAN-WAN connection.

The best approach in order to eliminate any SPoF is redundancy: by doubling (or even multiplying) any infrastructural element that could be a SPoF you can avoid a bad functioning state of the whole system in case of a failure.