Fault tolerant design is of paramount importance in Voice over IP. Any part that can move is a part that can break. VoIP and hosted PBX systems have a lot of moving parts, in both the figurative and literal sense. That’s why VoIP systems are designed not to break down when one part stops working.
Let’s start with a few definitions. “Uptime” is the measure of the time a machine has been working properly. The opposite of this is “downtime,” which, in computing terms, means a part or whole of something goes down, usually because of a glitch in the software, a power problem, or maintenance. “Redundant” in engineering terms, means to have more than the minimum of something in order to be more reliable. Redundancy is key to having as much uptime as possible.
When it comes to hardware, it is never a question of if something will break, it’s a question of when, and what you have already done to prepare for this eventuality. Before thinking of it in a grand scale, think of your own computer. Your hard drive spins, your CPU heats up and a fan cools it down, dust gets inside places you can’t reach, the metal fatigues and corrodes. Eventually, a part will break.
One way to prevent a failure is colocation (co-location). Colocation is to have several different companies’ servers in one location. Often, this location is in a remote, cold climate, which ensures physical security and prevents overheating. The servers can be “dedicated,” meaning one company owns specific parts, or they can be “shared,” meaning that many providers run on the same hardware. Providers pool their resources for the added computing power and bandwidth for the few times when demand is at its highest. Having enough bandwidth to handle the 1% of the time that you need it is a form of redundancy.
Another form of redundancy is to have many servers spread across a large area. Google, for example, has listed that they have six servers in the United States, one in Finland one in Belgium, one in Hong Kong, one in Singapore, and one in Taiwan. VoIP servers can also be spread across the world. RingCentral and 8x8 both have servers near both the East and West Coast of the United States, ensuring that downtime is minimal. When there is a natural disaster, often VoIP works better than traditional phone service because of these redundancies.
In Voice over Protocol, the data itself is also fault tolerant. First, the data itself is prioritized over other data, which is called Quality of Service (QoS). Second, the encoding allows for packets of data being lost by having more samples taken per second than necessary. Third, if a packet is lost, within nanoseconds, the receiving switch or IP phone can ask for another copy of that packet. It’s sort of like your phone asking a waiter to take back food that is cold.
In order to compete with traditional telephony, Voice over IP has to be just as reliable. Fortunately, due to constant innovation, VoIP systems are now just as reliable, if not more so, than their copper wire counterparts.