Network Latency Glossary

Time required for an API endpoint to process and return a request.

Maximum amount of data that can be transmitted across a network per second.

Amount of data that can be in flight on a network path, calculated as bandwidth × round-trip time.

The path segment with the least available capacity or highest contention that limits end-to-end throughput and often influences latency.

Excessive buffering that causes high latency and jitter.

Mechanisms used by protocols like TCP to prevent network overload.

Time required to establish a usable connection before application data exchange begins, including TCP and optionally TLS handshakes.

Distributed network of edge servers designed to reduce user latency.

Latency between servers located in different racks.

Network delay between systems within a data center.

TCP technique that delays acknowledgments to improve network efficiency.

Delay introduced by communication between distributed services.

Processing data closer to end users to reduce network delay.

Total time taken for a request to travel across the network and for the response to return.

Technique that distributes traffic across multiple network paths with equal routing cost, improving scale but sometimes affecting flow-level latency distribution.

Mechanism that signals impending congestion without dropping packets, allowing endpoints to reduce sending rates earlier.

Delay introduced when transmitting signals through fiber optic cables.

Technique regulating data transmission between sender and receiver.

Splitting large packets into smaller pieces for transmission.

Rate of useful application payload successfully delivered, excluding retransmissions, protocol headers, and overhead.

Traffic pattern where packets take an indirect path through an external network element or region and then return, often increasing latency unnecessarily.

Situation where one delayed packet blocks subsequent packets.

Network environment where packet delays are significantly high.

Number of network devices a packet traverses between source and destination.

Protocol used by diagnostic tools like ping to test connectivity.

High-performance networking technology widely used in HPC clusters.

Infrastructure where networks exchange traffic directly.

Common networking term for variation in packet delay over time; often used interchangeably with packet delay variation.

Increase in perceived latency caused by inefficiencies in system architecture.

Maximum allowable latency for a system to meet performance goals.

Distribution of latency measurements across requests.

Continuous measurement of network delay metrics.

Techniques used to reduce network delay across systems.

Statistical representation of latency distribution using percentiles.

Service-level agreement specifying maximum latency thresholds.

Sudden temporary increase in latency.

Application whose performance depends heavily on low network latency.

Switch connecting servers to the spine layer of a data center fabric.

Modern data center network topology designed to minimize latency between servers.

Rare but significant latency spikes affecting a small percentage of requests.

Network optimized to minimize packet transmission delays.

Largest TCP payload size that can be sent in a segment, typically derived from the path MTU minus IP and TCP header sizes. (Cloudflare Docs)

Largest packet size allowed on a network link.

Extremely small latency values measured in microseconds.

Latency added when services communicate across microservice architectures.

Latency measured in milliseconds, typical for internet-scale applications.

TCP optimization combining small packets to reduce overhead but sometimes increasing latency.

Temporary memory used to store packets before forwarding.

Latency involved in distributed operations across multiple nodes.

Condition where network demand exceeds available capacity, increasing latency.

High-speed interconnection architecture used in modern data centers.

Intermediate router or switch that forwards packets along the network path.

Hardware component connecting a system to a network.

Variation in packet latency over time.

The time it takes for a data packet to travel from a source to a destination across a network.

Basic unit of data transmitted across a network.

Sequence of routers and switches through which packets travel.

Direct interconnection between networks to reduce routing latency.

Mechanisms prioritizing certain traffic to maintain latency targets.

Determined path used by routing protocols to deliver packets.

Collection of performance metrics from network devices.

Layout of network devices and connections influencing latency.

Abstracting network resources to create flexible cloud networking environments.

Time required for a packet to travel from sender to receiver in a single direction.

Virtual network built on top of an underlying physical network.

Ratio of downstream to upstream bandwidth affecting congestion and latency.

Median latency experienced by requests.

Latency below which 95% of requests complete.

Latency below which 99% of requests complete.

Measurement describing variability in packet delivery times.

Situation where packets fail to reach their intended destination.

Amount of data contained in a network packet.

Process of discovering the maximum packet size that can traverse an end-to-end path without fragmentation. (Cloudflare)

Diagnostic tool used to measure network round-trip latency.

Time spent by routers and switches processing packet headers and routing decisions.

Time required for a signal to physically travel through a transmission medium such as fiber or copper.

Algorithm used by a network interface or operating system to schedule, shape, or prioritize queued packets.

Time packets spend waiting in network buffers before transmission.

Latency between servers located in the same rack.

Restricting network requests to maintain system stability.

Networking technology enabling direct memory access between servers with minimal latency.

Monitoring technique measuring latency experienced by real users.

Time required for a client request to reach a server and receive a response.

Total time between sending a request and receiving a complete response.

Time threshold before TCP resends unacknowledged packets.

RDMA implementation operating over Ethernet networks.

Total time required for a packet to travel from sender to receiver and back again.

Time required for routing systems to update paths after network changes.

Data structure used by routers to determine forwarding paths.

High network latency caused by long-distance satellite communication.

Time required to place all bits of a packet onto the link, which increases with packet size and decreases with link speed.

Infrastructure layer managing service-to-service communication.

Advanced NIC capable of offloading networking tasks from CPUs.

Network architecture allowing centralized traffic control.

Physical limit determining minimum latency over long distances.

Core switch connecting multiple leaf switches within a data center network.

Simulated tests used to measure expected network latency.

Queue management behavior that drops packets only when a buffer is full, often worsening latency and loss under congestion.

Slowest latency measurements occurring at high percentiles (e.g., P99).

Reliable transport protocol ensuring ordered delivery of packets.

Initial connection setup process used by TCP communications.

Technique that spaces out packet transmission over time to reduce burstiness, queue buildup, and latency spikes.

Process of resending packets that were lost in transit.

Initial phase of TCP congestion control gradually increasing transmission rate.

Parameter controlling how much data can be sent before acknowledgment.

Actual data transfer rate achieved over a network connection.

Time between sending a request and receiving the first byte of response data.

Latency added by negotiating a TLS session before protected application data can be exchanged.

Switch located within a server rack connecting servers to the network fabric.

Diagnostic tool that reveals each hop along a packet’s path.

Technique used to regulate network traffic to control latency and congestion.

Time required to place all packet bits onto a network link.

Lightweight transport protocol prioritizing speed over reliability.

Physical network infrastructure supporting overlay networks.