Gaming Networking Glossary: Essential Terms Explained

An ACK is a small confirmation packet telling the sender that data arrived.

Acknowledgements are how a connection-oriented protocol such as TCP knows whether data made it. If no ACK returns within the expected window, the sender retransmits. That safety net is excellent for downloads and terrible for real-time gameplay, which is why most games send positional updates over UDP and handle their own reliability at the application layer.

Anti-cheat is the software a game uses to detect tampering with the client or its network traffic.

Kernel-level anti-cheat inspects processes, memory and sometimes network adapters. Because route optimisers and VPNs insert a virtual adapter and rewrite the path your packets take, some anti-cheat systems flag or block them. Reputable gaming network tools publish compatibility lists; nothing that promises to modify game memory or packet contents is safe to run.

Example
A booster that only re-routes traffic is normally fine; a 'lag switch' style tool that drops or delays packets is not.

Anycast advertises one IP address from many locations so you reach the nearest instance automatically.

With anycast, dozens of servers around the world share a single address and internet routing decides which one you land on. It is widely used by DNS resolvers and CDNs, and increasingly by game login and matchmaking front ends. The 'nearest' instance is chosen by routing cost, not geography, so an anycast endpoint can occasionally send you somewhere illogical.

Application latency is the delay added by the game itself after packets have already arrived.

Once a packet reaches your machine, the game still has to parse it, apply it to the world state, simulate a frame and hand it to the GPU. That work happens on your CPU and is invisible to any network tool. When frame times spike, the game feels laggy even though ping, jitter and loss are all perfect.

Not the same as
Not the same as network latency — no booster, VPN or router can reduce it.

A backbone is a high-capacity long-distance network that carries traffic between cities, countries and continents.

Backbones are operated by tier-1 carriers, large ISPs, cloud providers and specialist routing companies. Your packets spend most of their journey on someone's backbone. The quality of the backbone your traffic lands on — how direct it is and how congested — largely determines the latency you cannot fix from home.

Bandwidth is the maximum rate at which data can move over a link, usually measured in megabits per second.

Bandwidth is capacity, not speed of arrival. A gigabit line and a 50 Mbps line deliver a single game packet at almost exactly the same moment; the gigabit line simply carries far more of them at once. Competitive games use tiny amounts of bandwidth — often under 1 Mbps — so upgrading a plan rarely improves how the game feels.

Example
A 128-tick shooter typically uses 200–600 Kbps. A 4K download saturating your line at the same time is what causes trouble, not the game's own bandwidth need.

Not the same as
Not the same as speed. Bandwidth is how wide the pipe is; latency is how long the trip takes.

Throttling is an ISP deliberately capping the rate of certain traffic, users or times of day.

Throttling can be plan-based (a fixed cap), congestion-based (slowing heavy users at peak), or protocol-based (deprioritising specific traffic types). Real throttling of game traffic is rare because games use so little bandwidth. What players usually experience as throttling is peak-hour congestion on a shared segment.

Not the same as
Not the same as congestion. Throttling is a policy; congestion is a capacity shortage.

A buffer is memory that holds packets temporarily while a device waits to send or process them.

Buffers exist in every router, switch, network adapter and game client. Small buffers absorb short bursts and are healthy. Oversized buffers hold packets far longer than the link needs, adding delay instead of dropping traffic — the cause of bufferbloat. Games also buffer received state deliberately, to smooth out jitter.

Bufferbloat is the latency spike caused when oversized router buffers fill up during heavy uploads or downloads.

When a link saturates, packets queue behind a backlog of already-buffered data. Idle ping of 20 ms can become 200 ms or worse the moment a cloud backup, console update or video call starts. The fix is not more bandwidth — it is smart queue management (SQM) on the router, which keeps queues short by design.

Example
Your ping doubles every time someone in the house uploads a video: classic bufferbloat, not an ISP fault.

CGNAT is an ISP sharing one public IP address across many customers, adding a second layer of address translation.

Because IPv4 addresses ran out, many ISPs — especially mobile and newer fibre providers — place customers behind carrier-grade NAT. Your router already performs NAT; CGNAT adds another translation upstream. It rarely adds meaningful latency, but it breaks port forwarding, can complicate peer-to-peer matchmaking and often forces games into relayed connections.

Not the same as
Not the same as ordinary NAT, which happens inside your own router and you control.

The client is the copy of the game running on your device, sending inputs to and receiving state from the server.

The client renders the world, predicts your movement locally so controls feel instant, and reconciles with the server when corrections arrive. Almost every visible network symptom — rubber-banding, teleporting opponents, shots that do not register — is the client and the server disagreeing about what happened and resolving that disagreement.

Cloud gaming runs the game on a remote server and streams video to you while sending your inputs back.

Cloud gaming inverts the usual requirements. Instead of tiny state packets, you receive a continuous high-bitrate video stream, so bandwidth genuinely matters — typically 15–45 Mbps. Latency matters even more than in native play, because every button press must travel to the datacentre and the resulting frame must travel back before you see it.

Not the same as
Not the same as native online play, where only game state crosses the network.

Congestion is what happens when more traffic arrives at a link than it can carry, causing queueing and eventually loss.

Congestion can occur on your own line, in your ISP's local segment, at an interconnect between networks, or at the game's own edge. Each location produces the same symptoms — rising latency, growing jitter and occasional packet loss — but requires a completely different fix, which is why localising it with a traceroute matters.

A connectionless protocol sends each packet independently, with no handshake and no guaranteed delivery.

UDP is the canonical example. There is no session setup, no ordering guarantee and no automatic retransmission — the application decides what to do about missing data. For games this is a feature: a positional update that arrives late is worthless, so it is better to discard it and use the next one.

A CDN is a distributed network of servers that caches content close to users.

In gaming, CDNs deliver patches, assets and store pages rather than live gameplay. A good CDN presence means fast downloads; it says nothing about the latency of the actual match server. Some publishers also run edge infrastructure for login and matchmaking, which is CDN-like but separate from the game servers themselves.

Cross-region matchmaking places you on a server outside your normal region, usually to fill a lobby faster.

When a playlist is thin — off-peak hours, a niche mode, a high rank — matchmakers widen the search radius. The result is a physically distant server and a latency floor no tool can remove. If your ping is normally 25 ms and occasionally 140 ms in the same game, region selection is the first thing to check.

A data center is the physical facility housing the game servers you connect to.

Game operators rent capacity in a handful of facilities per region. Your effective latency floor is set by the distance and route quality between your home and that specific building. Two 'EU West' servers can sit in different countries, which is why identical settings can produce different ping on different days.

A datagram is a self-contained packet that carries everything needed to route it, with no connection state.

UDP datagrams are the transport for nearly all real-time gameplay. Each one stands alone: it may arrive out of order, be duplicated, or never arrive. The game engine is responsible for sequencing and for deciding whether a late datagram is still useful — usually it is not.

A dedicated server is game-operator hardware in a datacentre that hosts the match, rather than one player's console or PC.

Dedicated servers give every player a consistent, well-connected host and remove the host-advantage problem. The alternative — peer-hosted or listen-server play — means one player's home connection determines everyone's experience, and that player enjoys near-zero latency while others do not.

Not the same as
Not the same as peer-to-peer hosting, where a player's own connection is the server.

The default gateway is the device your computer sends traffic to when the destination is outside your local network.

On a home network the default gateway is almost always your router, typically at an address like 192.168.1.1. Every packet bound for a game server passes through it first. If a traceroute shows high latency or loss on hop one, the problem is inside your house, not on the internet.

DNS translates human-readable names into the IP addresses machines actually connect to.

DNS runs before a connection exists. Your game client resolves a login or matchmaking hostname once, then talks to the resulting IP address directly for the rest of the session. A slow resolver can make launching, logging in or joining feel sluggish; it cannot make an established match laggy.

Not the same as
DNS lookup time is not in-game latency — switching resolvers will not lower your in-match ping.

Double NAT is address translation happening twice, usually because a router sits behind another router or modem.

The classic cause is an ISP gateway in router mode with your own router plugged into it. Double NAT rarely adds noticeable latency, but it produces strict NAT types, blocks port forwarding and can prevent voice chat or peer connections from establishing. The fix is putting one device into bridge or passthrough mode.

Download speed is how quickly data can be pulled from the internet to your device.

Download speed governs patch times and streaming quality. For live gameplay it is almost irrelevant beyond a low threshold — a competitive match needs a fraction of a megabit. Speed-test results are a poor proxy for gaming quality, because they measure capacity while ignoring latency, jitter and behaviour under load.

Not the same as
Not the same as low latency. Fast downloads and bad gaming performance regularly coexist.

Dynamic routing is routers exchanging live information so paths adapt automatically to failures and congestion.

The internet's inter-network routing protocol, BGP, is dynamic: paths change when links fail, capacity shifts or commercial agreements change. This is why your route to a game server can differ week to week without anything changing at your end, and why a booster's advantage on a given route is not permanent.

Not the same as
Not the same as a static route, which is configured by hand.

An edge server sits at the boundary of a provider's network, close to users, rather than deep in a central datacentre.

Route optimisers ingest your traffic at an edge node near you, carry it over their own backbone, and hand it off close to the game. CDNs use edge servers to cache patches. Proximity to an edge node is a large part of why a booster helps a lot for one player and not at all for another.

Encapsulation is wrapping a packet inside another packet so it can travel across a different network path.

Every tunnel — VPN or route optimiser — encapsulates your game packets inside its own transport. The wrapper adds header bytes, which reduces the payload space available inside the original MTU and can trigger fragmentation. Well-built tunnels account for this by lowering the effective MTU automatically.

Encryption scrambles packet contents so only the intended endpoint can read them.

Consumer VPNs encrypt everything, which costs a small amount of CPU time and adds header overhead. On modern hardware the processing cost is typically well under a millisecond, so encryption is rarely the reason a VPN raises ping — the longer physical path through the VPN server usually is. Most gaming route optimisers skip encryption deliberately.

Not the same as
Encryption overhead is not the main latency cost of a VPN; the detour is.

Ethernet is the wired local networking standard used by network cables between your device and router.

A wired Ethernet link gives you a dedicated, collision-free path with sub-millisecond, highly consistent latency. Wi-Fi shares airtime with every other device and neighbour on the channel, so it introduces retransmissions and jitter even when signal strength looks perfect. Cabling the gaming device is the single highest-value free fix available.

Fiber internet delivers data as light over glass strands, usually all the way to the premises.

Full fibre offers low, stable last-mile latency, symmetric or near-symmetric upload, and immunity to the noise and contention that affect copper and coaxial lines. It does not guarantee a good route: a fibre ISP with poor peering can still deliver worse game latency than a cable ISP with excellent interconnects.

Not the same as
'Fibre to the cabinet' is not full fibre — the last stretch is still copper.

A firewall filters traffic according to rules, allowing or blocking connections by address, port and protocol.

Firewalls exist in Windows, your router and sometimes at the ISP. For gaming, an overly strict firewall blocks inbound connections needed for voice chat or peer sessions, producing strict NAT warnings. A firewall adds negligible latency; the symptoms it causes are connectivity failures, not slowness.

FPS is how many frames your GPU renders and displays each second — a local rendering metric, not a network one.

FPS is produced entirely by your PC or console. It determines how smooth motion looks and how often your inputs can be sampled visually, but it has no bearing on how often the server updates the world. A 240 FPS machine on a 20-tick server still receives world state twenty times a second.

Not the same as
Not the same as tick rate. FPS is your machine's render rate; tick rate is the server's simulation rate.

Frame time is how long each individual frame takes to render, measured in milliseconds.

Average FPS hides inconsistency; frame time exposes it. A game averaging 144 FPS with occasional 30 ms frames feels stuttery, because the eye notices the outliers rather than the average. Frame-time spikes are a local hardware, driver or shader-compilation issue and are frequently mistaken for network lag.

Full tunneling routes all of your device's traffic through the VPN or optimiser.

Full tunnelling is the privacy-correct default for a VPN: nothing leaks outside the encrypted path. For gaming it is usually the wrong choice, because downloads, streaming and background sync then share the tunnel with your game packets and can congest it. Split tunnelling sends only the game through.

Not the same as
The opposite of split tunneling.

A game booster is a consumer product that claims to improve online game performance, most often by re-routing traffic.

The label covers two very different things. Genuine network boosters are route optimisers: they tunnel game traffic over a private backbone. 'System boosters' merely close background processes and free RAM, which does nothing for latency. Always establish which mechanism a product actually uses before judging its claims.

Not the same as
Not the same as a network route optimizer — some 'boosters' only tweak local system settings.

Netcode is the collection of techniques a game uses to keep many clients in agreement over an unreliable network.

Netcode covers tick rate, client prediction, interpolation, lag compensation and hit registration. It sets the ceiling on how good a game can feel online. Two titles played on the same connection with identical ping can feel completely different because their netcode makes different trade-offs between fairness and responsiveness.

The game server is the authoritative process that simulates the match and tells every client what happened.

It receives inputs, advances the simulation at a fixed tick rate, resolves conflicts and broadcasts state. Its location determines your latency floor; its tick rate determines the timing resolution of every action. Nothing you install locally can change either.

A gateway is any device that passes traffic between two different networks.

The term is used broadly. In a home, the gateway is usually the router — and often a combined modem-router supplied by the ISP. In a datacentre a gateway may be a dedicated appliance. The important idea is the boundary role: the point where your network ends and someone else's begins.

Not the same as
A router is a device; a gateway is a role. Your router is your gateway, but not every gateway is a home router.

Geo-routing directs your connection to an endpoint chosen by your apparent geographic location.

Matchmaking services, CDNs and login systems often route by geography inferred from your IP address. Using a VPN changes that inference, which is how players accidentally get matched into distant regions. Geo-routing decides where you go; it does not control how good the path is once chosen.

Hit registration is the server deciding whether a shot connected, based on the world state it believes was true.

Your client shows what it predicted; the server rules on what counted. Discrepancies come from tick resolution, interpolation delay, lag compensation windows and packet loss. Consistently poor hitreg with a good ping usually points to jitter or loss on the outbound path rather than to raw latency.

A hop is one router along the path between your device and the destination.

Each hop adds a small forwarding delay and a chance of queueing. Counting hops is a weak measure of route quality: a six-hop path that crosses an ocean is slower than a twelve-hop path within one country. What matters is where latency and loss actually appear, not how many devices are involved.

Not the same as
Hop count is not route quality. Fewer hops does not mean lower latency.

A host is any device with a network address — and, in matchmaking terms, the machine running the match.

In peer-hosted games one player's console or PC is the host, giving that player near-zero latency while everyone else pays the cost of reaching their home connection. In dedicated-server games the host is operator hardware, which is why dedicated hosting is fairer even when the average ping is higher.

ICMP is the control and diagnostic protocol behind ping and traceroute.

ICMP messages report errors and carry the echo requests used to measure round-trip time. Many routers deprioritise or rate-limit ICMP, so an intermediate hop showing 200 ms in a traceroute is often just a busy control plane, not a real problem. Only the final destination's numbers are reliably meaningful.

Input lag is the total delay between pressing a button and seeing the result on screen.

It is a chain: peripheral polling, USB stack, game input handling, simulation, render queue, driver, display processing and panel response. Every link is local. A network problem can add to what you perceive, but input lag measured properly exists even in a fully offline single-player game.

Not the same as
Not the same as network lag. Input lag persists offline; network lag does not.

An internet exchange is a shared physical facility where many networks interconnect and swap traffic directly.

Exchanges make peering cheap and local: instead of paying a carrier to haul traffic across a continent, two networks meet in the same building. Well-connected regions have large exchanges, which is why latency between two ISPs in the same city is usually excellent and why poorly connected regions are not.

Interpolation is the client smoothing between two received server states so motion looks continuous.

Because updates arrive only at the tick rate, the client deliberately renders slightly in the past and draws the in-between frames itself. This buys smoothness at the cost of a small, fixed delay. It is one reason a 20 ms ping does not mean you see the world 20 ms behind — the real figure includes interpolation delay.

An IP address is the numeric identifier used to route packets to a specific device or service.

Every packet carries a source and destination address. Your router holds one public address facing the internet and hands out private addresses inside the house. Games use your public address to route return traffic and, often, to infer your region for matchmaking.

IPv4 is the original 32-bit internet addressing scheme, offering about 4.3 billion addresses.

The address space exhausted years ago, which is why NAT and CGNAT are ubiquitous. Most games still use IPv4 by default. Its scarcity, not its performance, is the problem — an IPv4 path is not inherently slower than an IPv6 one.

IPv6 is the 128-bit successor to IPv4, giving every device a globally unique address.

IPv6 removes the need for address sharing, which can simplify peer connections and avoid CGNAT restrictions. Performance is usually comparable to IPv4, but occasionally an ISP's IPv6 routing is less mature and produces a worse path. If ping differs noticeably between the two, testing with IPv6 disabled is a legitimate diagnostic.

Your ISP is the company connecting your home to the internet and choosing how your traffic reaches everywhere else.

The ISP controls last-mile technology, local capacity, peering relationships and transit contracts. Those choices set your latency floor, your peak-hour stability and how directly you reach game datacentres. Two households on the same street with different ISPs routinely see materially different ping to the same server.

Jitter is the variation between consecutive latency measurements — how inconsistent your ping is.

A steady 60 ms connection feels better than one averaging 35 ms but swinging between 15 ms and 90 ms. Games buffer against small variation, but large or bursty jitter defeats interpolation and produces warping opponents and unreliable hit registration. Wi-Fi, bufferbloat and congested interconnects are the usual sources.

Not the same as
Not the same as packet loss. Jitter means packets arrive late and unevenly; loss means they never arrive.

A jitter buffer intentionally holds arriving data for a few milliseconds so uneven delivery can be played out smoothly.

Voice chat and streaming rely on jitter buffers; game engines use a conceptually similar interpolation delay. The buffer trades a little added latency for consistency. When jitter exceeds what the buffer was sized for, the smoothing fails and you see stutter or hear dropouts.

Lag compensation is the server rewinding the world to where you saw it when you fired, so high-latency players can still hit.

Without it, players with higher ping would need to lead every shot. With it, the shooter's view is favoured — which is why you sometimes die after already reaching cover on your own screen. Compensation windows are capped, so beyond a certain latency the advantage disappears entirely.

A LAN is the network inside one location — your router, its wired ports and its Wi-Fi.

LAN latency should be well under one millisecond over Ethernet. Anything higher, or any loss at all inside the LAN, points to a cable, switch, adapter or wireless problem you can fix yourself. Diagnosing the LAN first prevents wasted money on tools that can only affect the path beyond it.

The last mile is the final stretch of physical infrastructure between the ISP's local equipment and your home.

Its technology sets a floor you cannot beat: fibre adds around a millisecond, cable a few, DSL more, fixed wireless and satellite considerably more. Geostationary satellite adds hundreds of milliseconds by physics alone. No routing product can compensate for the last mile.

Latency is the time it takes for data to travel from one point to another, measured in milliseconds.

In gaming, latency is dominated by physical distance and the quality of the route, with queueing and wireless overhead on top. It is the metric that determines responsiveness. Bandwidth upgrades do not reduce it; shorter or better-managed paths do.

Not the same as
Latency is the general concept; ping is the specific round-trip measurement people quote.

Load balancing distributes connections across multiple servers or paths to avoid overloading any one of them.

Game operators balance matchmaking across server instances and datacentres; networks balance traffic across parallel links. Balancing keeps systems healthy, but it also means two sessions started minutes apart can land on different hardware or different routes, producing inconsistent ping with no change at your end.

Your local network is everything between your device and the point where your ISP's line leaves the property.

It includes the network adapter, cable or Wi-Fi link, any switches, and the router. Roughly half of all 'my internet is bad for gaming' complaints resolve inside this boundary — a wireless link, a bloated router queue, or another device saturating the upload.

Loss concealment is an application masking missing data rather than waiting for it to be resent.

Voice codecs synthesise a plausible fragment of audio; game engines extrapolate movement from the last known state. Concealment keeps things playable at low loss rates, which is why 0.5% loss may be invisible while 3% becomes obvious warping and desync. It hides the symptom, never the cause.

Matchmaking is the service that groups players into a lobby and assigns them to a server.

Matchmakers weigh skill, party size, queue time and estimated latency. When any of those constraints tighten — a small player pool, a high rank, an unusual mode — latency is normally the constraint that gets relaxed, and you are placed on a more distant server.

Mesh Wi-Fi uses several coordinated access points to cover a large area with one wireless network.

Mesh solves coverage, not latency. Wireless backhaul between nodes adds a hop and often halves airtime, so a device two hops from the router can see noticeably more jitter. For gaming, use a wired backhaul or, better, plug the gaming device into Ethernet on the nearest node.

Micro-stutter is brief, repeated hitching in rendered motion despite a healthy average frame rate.

Causes are local: shader compilation, background processes, thermal throttling, driver issues or a frame-pacing mismatch with the display. Because it looks similar to network warping, players often buy network tools to fix a rendering problem. Checking whether the stutter also occurs offline settles the question in seconds.

Not the same as
Not the same as network lag — micro-stutter appears in offline play too.

MTU is the largest packet size that can traverse a link without being fragmented.

Ethernet's standard MTU is 1500 bytes. Tunnels, PPPoE and some ISPs reduce the usable figure, so an oversized packet must be split or is silently dropped. Symptoms are odd and intermittent: logins that fail, patches that stall, sessions that drop while ping looks fine. Game packets are small, so pure gameplay is rarely affected.

Multi-path routing spreads traffic across several equal-cost paths at once.

Carriers use it for capacity and resilience. The side effect for real-time traffic is that consecutive packets may take different paths with slightly different delays, arriving out of order and adding jitter. Most implementations hash per-flow to keep a single session on one path, which largely avoids the problem.

NAT lets many devices share one public IP address by rewriting addresses and ports as traffic passes through the router.

NAT is why your console shows a 'NAT type'. Open types allow inbound connections and easy peer sessions; strict types rely on relays and can lengthen matchmaking or degrade voice chat. NAT itself adds negligible latency — its impact is on connectivity, not speed.

Not the same as
Not the same as CGNAT, which happens at the ISP and is outside your control.

The network adapter is the hardware in your device that connects it to the network, wired or wireless.

Adapters matter less than marketing suggests: a 'gaming' NIC does not lower latency across the internet. What does matter is driver quality, power-saving settings that idle the link, and offload features that occasionally misbehave. A failing or half-duplex link shows up as loss inside your own LAN.

Network congestion is sustained demand exceeding capacity somewhere along the path, degrading everyone's traffic.

It is usually time-of-day shaped: latency and jitter climb during evening peak and normalise overnight. If your idle-hours ping is excellent and your peak-hours ping is not, the constraint is capacity on a shared segment — your local ISP segment, an interconnect, or the game's own edge.

A network hop is a single router-to-router step in the journey your packets take.

Traceroute enumerates hops in order, showing where latency accumulates. Read the trend, not individual rows: latency should step up at long-distance links and then stay flat. A single high hop that does not affect later hops is a router deprioritising ICMP, not a fault.

A network route is the specific sequence of networks and routers your packets follow to a destination.

Routes are chosen by commercial agreements and routing policy, not by shortest distance. Traffic from London to a Frankfurt server can legitimately pass through Amsterdam. Route quality — directness, capacity and stability — explains most latency differences that hardware and bandwidth cannot.

A network tick is one cycle in which the server samples inputs, advances the simulation and sends updates.

Ticks are the heartbeat of an authoritative game. Everything that happens between them is resolved at the next one, which is why tick rate sets the timing resolution of the whole match. Your inputs are quantised to it regardless of how fast your machine runs.

A node is any addressable point on a network — a device, router, relay or server.

The word is deliberately generic. In route-optimiser marketing, 'nodes' usually mean the provider's ingress and egress points; in mesh Wi-Fi they mean access points. Ask which specific role is meant before drawing conclusions from a node count.

Online latency is the network component of the delay between your action and the server acting on it.

It excludes rendering, input handling and display delay, which are local. Separating the two is the whole diagnostic exercise: if online latency is low and stable but the game still feels sluggish, the remaining delay is on your machine or in the game's netcode.

Overhead is the extra bytes and processing added by protocol headers, encryption and tunnelling.

Every layer wraps your data in headers. A tunnelled, encrypted game packet may carry 60–100 bytes of wrapper around a payload of a few hundred. For low-bandwidth game traffic the bandwidth cost is trivial; the practical consequence is a reduced effective MTU rather than slower play.

A packet is the small unit of data the network moves, consisting of headers plus a payload.

A game sends a continuous stream of small packets: your inputs going out, world state coming back. Each is routed independently and may be delayed, reordered, duplicated or dropped. Every gaming network metric is ultimately a statement about what happened to these packets.

Packet capture records raw network traffic so it can be inspected in detail.

Tools such as Wireshark reveal which servers a game actually talks to, which ports and protocols it uses, and whether retransmissions or gaps are present. It is an advanced diagnostic and, because captures contain personal data, they should be handled carefully and never posted publicly unfiltered.

Packet duplication sends copies of each packet, so a single loss does not destroy the update.

Some route optimisers duplicate game packets across two paths and let the receiver take whichever copy arrives first. It genuinely masks sporadic loss and can trim jitter, at the cost of extra bandwidth. It cannot help when loss is caused by congestion on the shared last mile, because both copies queue behind the same backlog.

Packet loss is the percentage of packets that never arrive at their destination.

Loss is caused by congested queues, faulty cabling, wireless interference or overloaded equipment. Games handle a fraction of a percent invisibly; sustained loss above roughly 1% produces warping, missed hits and desync. Where the loss occurs matters more than the number itself — local loss you can fix, transit loss you cannot.

Not the same as
Not the same as jitter. Loss is data that never arrives; jitter is data that arrives at uneven times.

Prioritisation lets a device send some traffic ahead of the rest when the link is busy.

On a home router this is the useful half of QoS: mark game traffic as high priority so it jumps the queue ahead of uploads and backups. Prioritisation only works where you control the queue — your own upstream link. Beyond your router, other networks make their own decisions.

Retransmission is resending data that was not acknowledged, guaranteeing delivery at the cost of delay.

TCP retransmits automatically. For file transfers this is exactly right; for gameplay it is counterproductive, because a positional update resent 200 ms later describes a world that no longer exists. Games therefore use UDP and simply move on to the next update.

Pathping is a Windows tool that combines traceroute with sustained per-hop loss statistics.

Where a traceroute takes a snapshot, pathping sends many probes to every hop over a few minutes and reports loss percentages. That makes it far better at proving intermittent loss. Interpret it with the same caution as traceroute: only loss that persists through subsequent hops is real.

Peering is two networks connecting directly to exchange traffic between their own users, usually without payment.

Good peering means short, direct paths and stable latency. When your ISP peers with a game publisher's network or is present at the same exchange, packets take an obvious route. When it does not, they detour via a paid carrier — often through another city or country.

Not the same as
Not the same as transit. Peering exchanges traffic between two networks; transit buys access to everywhere.

Ping is the round-trip time for a small packet to reach a destination and come back, quoted in milliseconds.

Ping is both a tool and the number it produces. It measures the round trip, so it already includes the return journey — a common source of confusion. Ping alone is an incomplete picture: a stable 60 ms usually plays better than an average 35 ms with heavy variance.

Not the same as
Ping is a measurement of latency, not a synonym for it — and it measures the round trip, not one direction.

A port is a numbered endpoint that lets one device run many separate network services at once.

Games use specific UDP port ranges for gameplay and TCP ports for login and patching. Ports matter for firewall rules, NAT behaviour and quality-of-service rules that target game traffic. Publishers document their port ranges, which is what makes selective prioritisation practical.

Port forwarding tells your router to send inbound traffic on a given port to a specific device.

It is used to obtain an open NAT type for peer-hosted games and voice services. Port forwarding cannot lower latency — it only changes which connections can be established. It is also impossible behind CGNAT, because you do not control the address that inbound traffic arrives at.

A private backbone is a provider's own long-distance network, used to bypass congested public paths.

Route optimisers lease or operate backbone capacity between regions and carry your game traffic over it. The value depends entirely on whether their path is genuinely better than your ISP's default for your specific route. On well-peered domestic routes there is usually nothing to improve.

A protocol is the agreed set of rules governing how two systems format and exchange data.

Gaming touches several at once: IP for addressing, UDP or TCP for transport, ICMP for diagnostics, and the game's own application protocol on top. Knowing which layer a symptom belongs to is what stops you applying a transport-layer fix to an application-layer problem.

A public IP is the internet-routable address that identifies your connection to the outside world.

Your router presents one public address on behalf of every device in the house. Behind CGNAT you share that address with other customers. Games use it to route return traffic and often to infer your region, which is why a VPN can change which servers you are offered.

QoS is a router feature that classifies traffic and gives chosen types preferential treatment.

Effective QoS both prioritises game packets and caps the link slightly below its real capacity, so the queue stays with your router rather than the ISP's. Prioritisation without a rate limit achieves very little, which is why plain 'gaming mode' toggles often do nothing measurable.

A queue is the ordered line of packets waiting to be transmitted on a busy link.

Queueing is the main source of variable latency on the internet. A short queue costs a millisecond; a deep one costs hundreds. Every latency-management technique — SQM, QoS, prioritisation, congestion control — is ultimately about keeping queues short.

Queue management is the algorithm deciding which packets wait, which go first and which get dropped.

Modern algorithms such as fq_codel and CAKE drop or mark packets early to signal senders to slow down, keeping latency low under full load. This is the mechanism behind SQM and the reason a well-configured 100 Mbps line can feel better for gaming than an unmanaged gigabit one.

A relay is an intermediate server that forwards traffic between two endpoints that are not connecting directly.

Games use relays when NAT prevents a direct peer connection, and publishers increasingly relay all traffic through their own network for anti-DDoS and IP privacy. A relay adds a hop, but a well-placed one can shorten the effective path by putting traffic onto a better network sooner.

A relay node is a specific server in a provider's network where your traffic enters or leaves their path.

Route optimisers publish node locations because proximity determines benefit: if the nearest ingress node is 400 km away, your packets travel over the ordinary internet to reach it before any optimisation begins. Node coverage, not marketing claims, is the honest predictor of whether a booster will help you.

Retransmission is sending data again after it was lost or unacknowledged.

It occurs at several layers: Wi-Fi retransmits frames locally, and TCP retransmits segments end to end. Local wireless retransmission is a common hidden cause of jitter, because the delay it adds is invisible to any test that only measures loss.

Round-trip time is the total time for a packet to reach a destination and for the reply to return.

RTT is the technical name for what players call ping. It combines both directions, so a 60 ms RTT is roughly 30 ms each way if the path is symmetric — and paths often are not. Asymmetric routing is why upload-side problems can be invisible in a simple ping test.

A route is the path packets take through a series of networks to reach a destination.

Routes are decided hop by hop, and the return path may differ from the outbound one. Because routing follows commercial policy rather than geography, a route can be far longer than the straight-line distance — the gap that route optimisation exists to close.

Route optimisation forwards your game traffic over a provider's chosen path instead of your ISP's default route.

Traffic enters at a nearby node, crosses the provider's backbone, and exits close to the game datacentre. It helps most on long-distance or badly peered routes and does nothing when your default path is already direct. It cannot reduce the distance itself, fix your Wi-Fi, or change the server's tick rate.

Not the same as
Not the same as a VPN. A VPN optimises for privacy and location; an optimiser targets path latency and usually skips encryption.

A router forwards packets between networks and, at home, connects every device to your ISP's line.

Your router performs NAT, runs the firewall, manages the Wi-Fi and owns the outbound queue where bufferbloat happens. It is the single most influential piece of equipment you control — far more so than the network card in your PC.

Not the same as
A gateway is the role a router plays at a network boundary; the two words are often used interchangeably at home.

A routing table is the list of rules a router consults to decide where to send each packet next.

Internet routers hold hundreds of thousands of entries learned via BGP. Your PC has a small one too, which VPN and booster clients modify to capture game traffic. A misconfigured client that leaves stale routes behind after uninstalling is a classic cause of mysterious post-uninstall connectivity problems.

Rubber-banding is your character snapping back to a previous position after the server rejects your predicted movement.

Your client predicts movement locally; the server is authoritative. When your outbound updates are lost or badly delayed, the server never sees the movement and corrects you on the next reconciliation. It is nearly always an outbound loss or severe jitter symptom, not a hardware or frame-rate problem.

A server is a machine that provides a service to clients — in gaming, the match, matchmaking or login service.

Different servers do different jobs, and confusing them causes bad diagnosis. Slow logins point at authentication or DNS; laggy gameplay points at the match server's path. Pinging a publisher's website tells you nothing useful about the datacentre hosting your match.

A server-authoritative game treats the server's simulation as the only truth, overriding client claims.

This design prevents most cheating and guarantees a consistent world, at the cost of correction artefacts like rubber-banding when your updates go missing. Client-authoritative movement feels smoother but is trivially exploitable, which is why competitive titles all use server authority.

A server region is the geographic grouping of datacentres a game offers you.

Regions are broad. 'EU West' may span several countries, so two players in the same region can have very different latency. Where a game lets you pin a region or datacentre, testing each one manually is the cheapest latency improvement available.

Server tick rate is how many times per second the game server updates and broadcasts world state.

At 64 Hz the server resolves the world every 15.6 ms; at 128 Hz every 7.8 ms. That interval is a hard floor on timing precision for every player in the match, independent of ping or hardware. Only the operator can change it.

Not the same as
Not the same as your client frame rate — 240 FPS does not make a 20 Hz server update more often.

A session is one continuous logical connection between your client and a service.

A gaming session spans authentication, matchmaking and the match itself, often across several servers and protocols. Problems that appear only at session boundaries — joining, region switching, reconnecting — usually point at NAT, MTU or DNS rather than at latency.

Split tunnelling routes only selected applications through a tunnel and sends everything else out normally.

For gaming this is the correct configuration: the game takes the optimised or private path while streaming, downloads and browsing use the default route. It keeps bulk traffic from congesting the tunnel and avoids the tunnel's overhead on traffic that gains nothing from it.

Not the same as
The opposite of full tunneling.

SQM is a router configuration that keeps latency low under load by actively managing the outbound queue.

SQM sets a rate slightly below your real line speed and applies a modern queueing algorithm such as CAKE. Correctly configured, it holds ping within a few milliseconds of idle even during a full-speed upload. It is the definitive fix for bufferbloat and costs nothing beyond a capable router.

A static route is a manually configured path entry that does not change automatically.

Static routes are used on home and business routers to send specific destinations via a particular interface — for instance, forcing one console through a tunnel. They are precise but brittle: if the configured next hop fails, traffic simply stops instead of finding an alternative.

TCP is the reliable, ordered transport protocol that retransmits anything that goes missing.

TCP powers downloads, patching, logins and chat. Its reliability guarantees mean a single lost packet stalls everything behind it until the retransmission arrives — head-of-line blocking, which is unacceptable for real-time state. Games therefore use TCP for setup and UDP for gameplay.

Not the same as
Not interchangeable with UDP — reliability and low latency are competing goals.

Throughput is the data rate you actually achieve, as opposed to the capacity you are sold.

Throughput is bandwidth minus everything real: overhead, congestion, protocol behaviour and the slowest link in the chain. Gaming needs very little of it. It becomes relevant only when downloads or uploads saturate the link and start delaying the game's own packets.

Tick rate is the frequency, in hertz, at which the game server simulates and broadcasts the world.

It sets the timing granularity of everything in the match. On a 30 Hz server, two shots 20 ms apart may resolve in the same tick. Higher tick rates cost the operator money, so it is a design and budget decision rather than something a player can influence.

Not the same as
Not the same as FPS. Tick rate is server-side simulation; FPS is client-side rendering.

TTL is a counter in each packet that decreases at every hop and causes the packet to be discarded at zero.

TTL prevents packets circulating forever in a routing loop. Traceroute exploits it deliberately: it sends packets with TTL 1, 2, 3 and so on, and each expiring packet triggers an error message that reveals the router at that position in the path.

Traceroute maps every router between you and a destination, reporting the latency to each one.

It is the tool that localises a problem: inside your home, on your ISP, at an interconnect, or near the game. Read it for trends. Latency that jumps at one hop and stays elevated afterwards is real; a single high hop followed by normal ones is a router deprioritising diagnostic traffic.

A traffic relay is infrastructure that receives your traffic and forwards it on, hiding the true endpoint.

Publishers relay gameplay to shield server addresses from DDoS attacks and to protect player IPs. This is why pinging the address you see in a network monitor may measure the relay rather than the match server, and why some route optimisers cannot improve certain titles at all.

Traffic shaping smooths traffic to a defined rate by delaying packets rather than dropping them.

Shaping is used by ISPs to enforce plan limits and by home routers as part of SQM. Shaping applied thoughtfully keeps latency low; shaping applied bluntly to a congested link simply adds queueing delay. The difference lies entirely in the queueing algorithm behind it.

Transit is a paid service in which one network carries another network's traffic to the rest of the internet.

Where peering is a direct, usually settlement-free exchange, transit is a commercial contract for reachability. ISPs buy transit for destinations they do not peer with. Because transit costs money per megabit, an ISP's cheapest route is often not the shortest — the structural reason routes can look illogical.

Not the same as
Not the same as peering. Transit buys access to everywhere; peering exchanges traffic with one partner.

A tunnel encapsulates your traffic inside another connection so it travels a path of the provider's choosing.

Both VPNs and route optimisers build tunnels; they differ in what they optimise for. A tunnel adds header overhead and an ingress hop, so it must deliver a genuinely better path to be worth it. On an already-direct route, a tunnel almost always makes things marginally worse.

UDP is the lightweight, connectionless transport protocol used for nearly all real-time gameplay.

UDP sends datagrams with no handshake, no ordering and no retransmission. That is precisely what a game wants: stale data is worthless, so it is better to drop it and use the next update. Reliability that genuinely matters is implemented selectively by the game itself.

Not the same as
Not a 'worse' TCP — different guarantees for different jobs.

Upload congestion is your outbound link being saturated, delaying the small packets your game needs to send.

This is one of the most common gaming problems and has nothing to do with how much upload capacity you have. A cloud backup or game update will happily fill any link, and your game's packets then queue behind it. SQM or prioritisation fixes it; a bigger plan usually does not.

Not the same as
Not the same as low upload speed. A gigabit upload still congests if something saturates it.

Upload speed is the rate at which your connection can send data outward.

Games need very little upload — inputs are tiny. Upload matters because asymmetric connections such as older cable and DSL have small upstream capacity that is easily saturated, and because your outbound packets are the ones the server needs in order to accept your movement and shots.

An upstream provider is the larger network your ISP buys connectivity from.

Small and regional ISPs depend on one or more upstreams to reach the wider internet. If the upstream has poor capacity toward a game's network, your ISP's own equipment can be flawless and your latency still poor. This dependency is invisible from a speed test and visible in a traceroute.

A VPN encrypts your traffic and routes it through a server you choose, primarily for privacy and location flexibility.

VPNs are built for confidentiality and access, not latency. Routing through an extra server usually adds distance and therefore ping. A VPN can occasionally help — bypassing a badly congested ISP path or ISP-level throttling — but that is a side effect, not the design goal.

Not the same as
Not the same as a game booster or route optimiser, which target path latency and typically skip encryption.

A VPN server is the endpoint where your encrypted tunnel terminates and your traffic rejoins the ordinary internet.

Its location determines both your apparent location and the shape of your path. A VPN server between you and the game can shorten a bad route; one in the opposite direction lengthens every journey. Choosing a server close to the game datacentre is the only configuration that stands a chance of helping latency.

A WAN is a network spanning a large geographic area — on a home router, the WAN port faces the internet.

Everything past the WAN port is outside your control. That boundary is the most useful diagnostic line in home networking: problems on the LAN side you can fix yourself, problems on the WAN side need your ISP or a routing product, and confirming which side is at fault comes first.

A Wi-Fi channel is the specific frequency slice your wireless network transmits on.

Neighbouring networks on the same or overlapping channels share airtime, causing retransmissions and jitter. The 2.4 GHz band has only three non-overlapping channels and is heavily congested; 5 GHz and 6 GHz have far more room. Choosing a clear channel is a real, if modest, improvement for wireless gaming.

Wi-Fi interference is competing radio traffic and noise degrading your wireless link's reliability.

Interference causes frames to be retransmitted at the radio layer. Those retries add milliseconds invisibly, so the connection shows full signal bars while producing significant jitter and occasional loss. It is the most common cause of gaming problems that no internet-side tool can fix.

Example
Ping is stable at 3 a.m. and erratic every evening on the same connection — usually neighbouring networks, not your ISP.

A wired connection uses an Ethernet cable rather than radio, giving consistent, near-zero local latency.

Cabling removes airtime contention, interference and wireless retransmission in one step. It is the first recommendation in every credible latency guide and costs less than any software subscription. Where cabling is impossible, powerline or MoCA adapters are usually better than Wi-Fi for a fixed gaming device.