{
  "id": "circuit-switching",
  "title": "Circuit Switching",
  "lang": "en",
  "source_lang": "en",
  "summary": "Circuit switching establishes a dedicated communication channel between endpoints, reserving network resources for the duration of a connection.",
  "category": "computer-science",
  "subcategory": "networks-security",
  "type": "concept",
  "aliases": [
    "Circuit-switched communication"
  ],
  "generated_at": "2026-11-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/circuit-switching/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/circuit-switching/",
      "json": "https://aiwiki.page/zh/circuit-switching.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Circuit switching is a method of telecommunications in which a network establishes a dedicated communication channel between two endpoints and reserves the resources needed to maintain it for the duration of a connection. It is historically associated with telephone networks, where switches connect successive transmission links to carry a conversation. Unlike packet switching, which shares transmission capacity among individually forwarded packets, circuit switching assigns capacity to a connection even when the users are temporarily silent or sending no useful information.\n\nA circuit need not be a separate physical wire running between the participants. It can consist of allocated time slots, frequency bands, optical wavelengths, or other resources across shared transmission facilities. The defining feature is the reservation of an end-to-end channel, not the physical form of the medium."
    },
    {
      "heading": "Operating principle",
      "text": "A circuit-switched connection normally has three phases:\n\n- Connection establishment. The network identifies the destination, selects a route, and allocates resources along that route. If sufficient resources are unavailable, the connection attempt may be blocked.\n- Information transfer. The endpoints communicate through the established channel. The network maintains the resource allocation throughout the connection, including idle periods.\n- Connection release. When the communication ends, the network clears the connection and makes its resources available to other users.\n\nIn a conventional telephone call, establishment also involves actions such as checking the destination line, ringing the called telephone, and detecting an answer. The details and timing of resource allocation depend on the network and its signaling procedures.\n\nOnce established, a circuit generally remains on its selected path until release or a network reconfiguration. Information does not need a destination address attached to every individual unit of user data: the switches already hold the connection state needed to direct it.\n\nThe reservation applies to a particular capacity allocation, not necessarily to an entire transmission link. A trunk carrying many simultaneous calls can allocate a separate channel to each one."
    },
    {
      "heading": "Physical and logical implementations",
      "text": "Circuit switching can operate with either analog or digital signals. It should not be equated with analog telephony.\n\nSpace-division switching\n\nIn space-division switching, a switch establishes a physical signal path between an input and an output. Manual switchboards, electromechanical switching systems, and electronic switching matrices can all implement this principle.\n\nLarge switches often use multistage arrangements rather than a separate switching element for every possible input–output pair. Depending on their design and current allocation of internal paths, these arrangements may be blocking or nonblocking. A blocking switch can fail to establish a requested connection even when the desired output is free, because a suitable internal path is unavailable.\n\nFrequency-division and time-division channels\n\nThrough multiplexing, many circuits can share one physical medium.\n\nIn frequency-division multiplexing, separate channels occupy different frequency bands. Analog telephone carrier systems used this technique to combine multiple conversations on shared trunks.\n\nIn time-division multiplexing, channels occupy recurring time slots in a repeating frame. A digital telephone connection can receive a fixed slot on each link along its route. Digital switches may transfer the contents of an incoming slot into a different outgoing slot, so a circuit does not have to use the same slot number throughout the network.\n\nA common digital voice channel carries 8,000 samples per second, with eight bits per sample, producing a nominal rate of \\(64{,}000\\) bits per second. This rate describes the voice channel itself; framing, signaling, and other overhead can require additional transmission capacity.\n\nOptical circuits\n\nCircuit-oriented connections can also be established through optical networks. With wavelength-division multiplexing, multiple optical channels share a fiber at different wavelengths. An optical connection may reserve a wavelength or another defined portion of optical transmission capacity across a route.\n\nSuch connections can carry aggregated traffic, including traffic originating in packet networks. Circuit switching and packet switching can therefore coexist at different layers of the same communications system."
    },
    {
      "heading": "Signaling and connection control",
      "text": "Signaling is the exchange of control information used to establish, supervise, modify, and release connections. It is distinct from the conversation or other user information carried by the connection.\n\nHistorically, signaling could travel within the same channel used for speech. Later systems increasingly separated control traffic from user traffic. Signaling System No. 7, for example, provides a signaling framework widely used in telephone networks.\n\nThis separation illustrates an important distinction: a circuit-switched service can use a packet-based signaling network. The classification of the user channel does not require its control messages to use the same switching method.\n\nConnection control also performs admission control. Rather than accept every request and allow all active connections to compete for insufficient capacity, a circuit-switched network can reject a new request when the required resources are unavailable. Once admitted, a connection normally retains its allocation."
    },
    {
      "heading": "Performance and capacity",
      "text": "Predictable service after setup\n\nAn established digital circuit provides a defined transmission rate and generally predictable delay. Because its capacity is reserved, unrelated traffic does not ordinarily introduce the varying output-queue delays typical of shared packet links.\n\nCircuit switching does not eliminate all delay. Signals still incur propagation delay, and equipment can introduce encoding, switching, synchronization, and other processing delays. Connection establishment also takes time before communication can begin.\n\nIdle capacity and utilization\n\nReserved resources remain unavailable to other circuits even when an established connection carries no useful information. This is well suited to sustained traffic with predictable rate requirements, but it can be inefficient for bursty activity.\n\nFor example, a terminal may require capacity briefly to send a command and then remain idle while waiting for a response. A circuit reserved throughout that interaction continues to occupy network resources during the pauses. Packet switching can instead allocate transmission opportunities to whichever users currently have packets to send.\n\nFor a circuit of capacity \\(R\\), if useful traffic averages \\(r\\) over its holding time, its useful-load fraction is\n\n\\[u=\\frac{r}{R}.\\]\n\nA low value of \\(u\\) indicates that much of the reserved capacity is unused. This simple measure does not include signaling overhead, physical-layer overhead, or network-wide effects.\n\nBlocking and traffic engineering\n\nA network with finitely many circuits cannot accommodate an unlimited number of simultaneous connections. Traffic engineering estimates how much capacity is needed to meet a target blocking probability.\n\nOffered traffic is commonly measured in erlangs. For a simple traffic model, the offered load is\n\n\\[A=\\lambda h,\\]\n\nwhere \\(\\lambda\\) is the average rate of connection attempts and \\(h\\) is the mean holding time, expressed in consistent time units. One erlang corresponds to an average demand equivalent to one continuously occupied circuit.\n\nUnder specific assumptions—including independent Poisson arrivals, a fixed pool of identical circuits, and blocked calls leaving without waiting—the Erlang B formula gives the blocking probability. Actual networks may require more elaborate models to account for retries, alternative routing, correlated demand, or different service classes."
    },
    {
      "heading": "Comparison with packet switching",
      "text": "The central difference is how capacity is allocated.\n\nFeature | Circuit switching | Packet switching\nResource allocation | Capacity reserved for an established connection | Capacity generally shared among packets\nSetup | Normally required before user transfer | Not inherently required, though some packet services establish connections\nAddressing | Connection state determines the established channel | Packets carry addressing or forwarding information\nIdle periods | Reserved capacity remains allocated | Other traffic can generally use available capacity\nCongestion | Can block new connections | Can produce queues, delay, packet loss, or admission rejection\nService behavior | Predictable allocated rate after establishment | Depends on traffic, scheduling, and any resource guarantees\n\nThese are characteristic differences rather than absolute rules. Packet networks can reserve resources and provide quality-of-service guarantees. Conversely, a circuit-switched connection can fail because of equipment faults or loss of a transmission link.\n\nA virtual circuit in a packet network is not necessarily a dedicated circuit. It establishes a logical association and forwarding state, but packets belonging to different virtual circuits may still compete for shared bandwidth. Likewise, a connection-oriented transport protocol such as TCP does not itself reserve a circuit across the Internet."
    },
    {
      "heading": "Historical development and applications",
      "text": "Circuit switching developed with telephone networks. Early exchanges used operators to connect subscriber lines manually. Automatic electromechanical switches subsequently performed connection establishment without requiring an operator for each call. Electronic control and digital switching later expanded the scale and flexibility of telephone exchanges.\n\nThe public switched telephone network became the principal large-scale application of circuit switching. Circuit-switched facilities also carried data through modems and digital services such as the Integrated Services Digital Network.\n\nAs computer communication expanded, bursty data traffic encouraged the adoption of packet switching. Packet-based voice systems subsequently allowed telephone conversations to share networks with other kinds of traffic. Modern voice services can therefore present a familiar telephone interface without establishing an end-to-end circuit-switched channel.\n\nDedicated optical connections and other circuit-oriented transport facilities remain useful for carrying stable, high-capacity traffic. A packet-based service may traverse such facilities, while gateways can interconnect packet voice systems with circuit-switched telephone networks. Whether a service is circuit-switched consequently depends on the network layer and portion of the connection being considered."
    }
  ],
  "links": [
    "telephone",
    "packet-switching",
    "multiplexing",
    "frequency-division-multiplexing",
    "bit",
    "wavelength-division-multiplexing",
    "internet"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}