Network Latency

Network latency is the delay between the moment data is sent from one place on a network and the moment it arrives at its destination. It is usually measured in milliseconds, either one way or as round-trip time, the time for a message to reach its destination and for a reply to come back. Latency is different from bandwidth, which is how much data a connection can carry per second: a link can have high bandwidth and still be slow to respond.

Latency affects almost every digital financial service. It shapes how quickly a card payment is approved, how fast an online banking page loads, whether a trader's order reaches an exchange before a rival's, and how quickly a new block spreads across a blockchain. In electronic trading, firms spend heavily to cut latency by microseconds, while in payments and crypto the main concern is keeping delays low and predictable.

Imagine shouting a question across a valley and waiting for the echo. The time it takes for the sound to reach the far side and come back is like latency. A wider valley means a longer wait, and wind or obstacles can add more delay. However loudly you shout, you cannot make the sound travel faster than nature allows.

On the internet, the "valley" is the physical distance and the equipment between two computers. Data travels through cables at close to the speed of light, but it also stops at routers along the way, waits in queues when lines are busy and gets processed at each end. All of those small waits add up to the delay a user experiences.

• Propagation delay. The time for signals to travel the physical distance. Light in optical fibre travels at roughly two-thirds of its speed in a vacuum, so every 1,000 kilometres adds about 5 milliseconds one way.
• Transmission delay. The time needed to push all the bits of a message onto the link, which depends on message size and bandwidth.
• Processing delay. Time spent by routers, firewalls and servers inspecting and forwarding data.
• Queuing delay. Waiting time when links or devices are congested, which can vary sharply from moment to moment and causes "jitter".

Latency is commonly measured with simple tools that send a small message to an IP address and time the reply, and with specialised monitoring in trading and payment systems.

In electronic markets, the first order to reach an exchange's matching engine gets the best price, so speed is a competitive advantage. High-frequency trading firms place their servers in the same data centres as exchange systems, an arrangement called co-location, and build dedicated links between financial centres. Around 2010 a new fibre-optic cable was laid on a straighter route between Chicago and New York to shave milliseconds off trading times, and it was later outpaced by microwave and millimetre-wave radio links, which travel through air faster than light moves through glass. Speed advantages also matter for arbitrage between venues. Critics argue that the race for lower latency adds cost without improving markets, and some exchanges have introduced small deliberate delays, known as speed bumps, to blunt it.

Card payments involve several hops, from the terminal to the merchant's acquirer, the card network and the issuing bank, and back, usually within a couple of seconds, a process described under authorization and settlement. Real-time payment systems set strict limits on how long a transfer may take end to end, so participating banks must keep latency low. High or unpredictable latency leads to timeouts, duplicate payment attempts and poor customer experience, so payment firms monitor it closely and route traffic through data centres close to their users, often using cloud computing regions.

Public blockchains rely on nodes spread around the world passing messages to each other through peer-to-peer networks, often built with tools such as libp2p. When a new block is produced, it takes time to reach every node. If two miners or validators produce blocks before hearing of each other's, one may be discarded as a stale block, wasting work. This is one reason networks cannot simply shorten their block time indefinitely: the shorter the interval, the larger the share of time lost to propagation delay. Latency also creates advantages for well-connected participants. Validators and searchers located close to major relays and builders can react to new information faster, a factor in competition for maximal extractable value, and critics worry this encourages geographic concentration of nodes in a few data centres. Protocol designers address these issues through faster propagation methods, compact block formats and consensus rules that tolerate some delay.