Components of Latency

Latency (in computer networking) is the amount of time that passes between connected computer's sending an instruction to a server and its receiving of a response from it. In an MMO like World of Warcraft, it greatly affects how much a player enjoys the game. In a raiding scenario, a delay of more than 500ms (half of one second) is usually considered unacceptable.

While network latency is a significant component of the total delay, it is not the only one. In any case where signals need to be sent from one physical place to another, a time delay will exist, no matter how short the link or how fast the signals can travel.

The total delay relevant to playing an online game can be defined as:

The amount of time that passes between issuing an instruction on an input device and realising the response that is returned.

Then:

Total delay = Network latency + Computer latency + Reaction time

Network latency
Network latency is generated from the computer network as the instruction traverses it aboard a data packet. This happens because (like with any moving object), it does not have an infinite speed and therefore will take some amount of time to reach the destination. Were the link between the client and the server a direct one, the only latency would be from the movement of the instruction across physical space. These delays are negligible anyway (optical fibre packets, for example, travel at the speed of light of 299 792km/s).

However, since all WoW clients connect to WoW servers over a packet-switched Internet, there will be further delays generated by the routers in between links. These routers often need to transfer the data packet from one physical medium to another (e.g. electrical cabling to optical fibre), as well as direct packets along optimal paths, which will take time as the router processes the packets coming in. The routers also maintain buffers of incoming and outgoing packets, which may fill up under network congestion. A packet-switched link over the same distance would likely introduce more delay than a direct link.

Packet switching allows this many nodes to connect to the Internet.

Computer latency
Computer latency is generated when both the client and server computers handle the instructions and responses. This may involve:

• Processing the instruction/response with the CPU or any device's processing unit
• Transporting the instruction/response along one of the computer's many buses
• Storing and recalling the instruction/response from memory
• In the case of the client, sending the response from the video card to the display's surface

In the current age of powerful microcomputers, the first three are negligible for the client (for example, CAS latency in the RAM modules is measured in the order of nanoseconds). However, the fourth one is in the order of milliseconds and may become a material concern for some players. Typical displays have a 5ms response time, but better displays can display as fast as 3ms.

For the server, computer latency only becomes noticeable as it approaches capacity. At this weight of load, the server risks falling behind and accumulating a backlog of unprocessed instructions.

Reaction time
Reaction time is generated as the player's body handles the response received from the display and/or speakers. Average reaction time for visual stimulus is near 190ms. Mirroring what happens in a computer, this may involve:

• Processing the response in the brain's neural network
• Transporting the response or any derivative reactions along neurons
• Storing and recalling the response in the brain's memory

The quest to minimise latency
Because the total delay becomes more undesirable as it increases, players strive to minimise it. They may do so by:

• Obtaining a good connection to the server(s)
• Investing in more powerful computer equipment
• Training their reaction time
• Asking the game host (for WoW US, it is Blizzard Entertainment) to improve their server hardware and connections