Computer NetworksNetwork Layer Services

Network Layer Services

LevelIntermediate

Duration75 mins

TopicNetwork Layer Services

2 / 5

Connection-Oriented Service

Establishing a Path Before Communication

When you make a phone call, the telephone network doesn't just send your voice as independent packets hoping they arrive. Instead, a connection is established first—a path through the network is set up, switches are configured, and resources are reserved. Only then does conversation begin, with every sound byte traveling the same predetermined path in order.

This is connection-oriented service: before any data flows, both endpoints agree to communicate, a path is established through the network, and possibly resources are reserved along that path. The network maintains state about the connection, ensuring all data follows the established route.

While the Internet's IP layer is famously connectionless, connection-oriented service has played—and continues to play—crucial roles in networking. Understanding this paradigm is essential for comprehending technologies like ATM, MPLS, traditional telephony, and why certain applications require connection-like guarantees.

What You Will Learn

By the end of this page, you will understand connection-oriented service in depth—the three phases of connection lifecycle (establishment, data transfer, termination), virtual circuit technologies, label switching mechanisms, the tradeoffs between connection-oriented and connectionless approaches, and when each model is appropriate.

Defining Connection-Oriented Service

Connection-oriented service requires that communicating parties establish an explicit contract (connection) before data transfer begins. The network creates a logical path, and all data belonging to that connection follows the same path.

The Core Definition:

A connection-oriented network service operates through three distinct phases:

Connection Establishment — A signaling process sets up the path through the network. Intermediate switches/routers are configured with forwarding information for this specific connection.
Data Transfer — Data flows along the established path. All packets follow the same route in sequence.
Connection Termination — When communication ends, a teardown process releases resources and removes connection state from network devices.

Key Characteristics:

Fundamental Properties

•Explicit Setup Phase — Before data transfer, signaling messages traverse the network to establish the connection and configure intermediate devices.
•Stateful Network Devices — Switches/routers maintain per-connection state: connection identifiers, forwarding entries, possibly reserved resources.
•Fixed Path — All packets of a connection follow the same path, ensuring in-order delivery (barring errors).
•Connection Identifier — Instead of full destination addresses, packets carry short connection identifiers (labels) that switches use for fast forwarding.
•Guaranteed Ordering — Because packets follow the same path, they arrive in the order sent (unless lost).
•Possible Resource Reservation — The connection setup can reserve bandwidth, buffer space, or processing capacity along the path.

Virtual Circuits vs. Physical Circuits

Early telephone networks used physical circuit switching—electrical paths were literally connected. Modern connection-oriented networks use 'virtual circuits'—the path is logical, with packets sharing physical links but maintaining separate connection state. The circuit is 'virtual' because no dedicated physical wire exists; it's a software abstraction.

The Virtual Circuit Model

A virtual circuit (VC) is the fundamental abstraction of connection-oriented networking. It represents a logical path through the network that behaves like a dedicated connection.

How Virtual Circuits Work:

1. Connection Request: The source sends a setup message containing the destination address. This message travels through the network, with each switch:

Allocating local resources (buffer space, table entries)
Assigning a Virtual Circuit Identifier (VCI) for this connection on each link
Adding an entry to its forwarding table: (incoming port, incoming VCI) → (outgoing port, outgoing VCI)
Forwarding the setup message toward the destination

2. Connection Confirmation: When the setup reaches the destination, a confirmation travels back, finalizing the path. Now all switches along the route have forwarding entries for this VC.

3. Data Transfer: Packets carry only the VCI (not full addresses). Each switch:

Looks up (incoming port, VCI) in its table
Finds (outgoing port, new VCI)
Replaces the VCI with the new value
Forwards the packet

4. Teardown: When communication ends, teardown messages traverse the path, and switches delete their VC entries.

Converting Mermaid diagram...

Virtual Circuit Identifier (VCI):

The VCI is a local label significant only on a single link. As a packet traverses the network, its VCI changes at each switch—this is called label swapping. The same VC might use VCI 5 on one link and VCI 847 on the next.

Why local labels?

Labels can be short (typically 16-32 bits vs. 32+ for addresses)
Fast table lookup (labels can be array indices)
Same label can be reused on different links for different VCs

Example Switch Forwarding Table (Virtual Circuit)
Incoming Port	Incoming VCI	Outgoing Port	Outgoing VCI
Port 0	12	Port 2	41
Port 0	25	Port 1	18
Port 1	33	Port 2	55
Port 2	87	Port 0	12

Types of Virtual Circuits

Virtual circuits come in two primary forms, each suited to different usage patterns.

Switched Virtual Circuits (SVCs):

SVCs are dynamically established on demand, like phone calls. When communication is needed, the source initiates connection setup. When finished, the connection is torn down.

Characteristics:

Established per-session
Setup and teardown overhead for each connection
Resources may be allocated only for the connection duration
Flexible—no pre-configuration needed
Used for sporadic, on-demand communication

Permanent Virtual Circuits (PVCs):

PVCs are pre-configured connections that remain established indefinitely. They're set up administratively (not by signaling) and stay active continuously.

Characteristics:

Configured by network administrators
No per-session setup delay
Resources permanently allocated
Lower flexibility—changes require reconfiguration
Ideal for constant communication between fixed endpoints
Common in enterprise WAN connections (e.g., between branch offices)

Switched Virtual Circuits

•Established dynamically via signaling
•Setup time before data can flow
•Resources allocated per-session
•Automatic teardown when done
•Scales well for many destinations
•Phone calls, video conferences

Permanent Virtual Circuits

•Pre-configured administratively
•No setup delay—always ready
•Resources permanently reserved
•Changes require manual config
•Efficient for fixed endpoints
•Branch office connections

Hybrid Approaches

Some networks use 'Soft PVCs'—connections that are administratively defined but automatically re-established if they fail. This combines PVC simplicity with some SVC resilience. Modern MPLS networks often use this approach for important traffic classes.

Technologies Using Connection-Oriented Service

Several major networking technologies employ connection-oriented service principles. Understanding these examples illustrates where and why this model is used.

ATM (Asynchronous Transfer Mode):

ATM was a comprehensive connection-oriented technology designed in the 1990s to unify voice, video, and data networking.

Key Features:

Fixed-size 53-byte cells (48 payload + 5 header)
Virtual Paths (VP) and Virtual Channels (VC)
Hardware-based cell switching
Built-in QoS with multiple service categories
Signaling protocol for SVC establishment

ATM saw widespread deployment in carrier networks but lost to IP-based solutions for end-to-end communication due to complexity and overhead.

MPLS (Multiprotocol Label Switching):

MPLS takes the best of both worlds—connectionless IP for universal addressing with label-switching for efficient forwarding and traffic engineering.

Key Features:

Labels inserted between Layer 2 and Layer 3 headers
Label switching provides connection-like forwarding
Label Distribution Protocol (LDP) or RSVP-TE for path setup
Traffic engineering enables explicit path control
Widely deployed in carrier and enterprise networks

MPLS is sometimes called 'Layer 2.5' because it operates between traditional Layer 2 and Layer 3.

mpls_label_structure.txt
MPLS Label Stack Entry (32 bits):
+---------------+---+---+--------+
|     Label     |TC |S  |  TTL   |
+---------------+---+---+--------+
|   20 bits     |3b |1b | 8 bits |
 
Label: Virtual circuit identifier (0-1,048,575)
TC: Traffic Class (QoS marking)
S: Bottom of Stack flag (1 = last label)
TTL: Time To Live (prevents loops)
 
MPLS Packet Structure:
+==============+==============+==============+
| Layer 2 Hdr  | MPLS Labels  | IP Packet    |
| (Ethernet)   | (1 or more)  | (original)   |
+==============+==============+==============+

Frame Relay:

Frame Relay was a connection-oriented WAN technology popular in the 1990s and 2000s.

Key Features:

Data Link Connection Identifier (DLCI) for virtual circuits
Variable-length frames (more efficient than ATM cells)
Primarily PVCs (SVCs were defined but rarely used)
Committed Information Rate (CIR) for bandwidth guarantees
Largely replaced by MPLS and Ethernet services

Traditional Telephony (PSTN):

The Public Switched Telephone Network exemplifies connection-oriented design:

Call setup establishes a circuit
64 kbps reserved per call direction
All voice samples follow the same path
Call teardown releases resources

Connection Establishment in Detail

The connection establishment phase is critical—it's where paths are computed, resources are reserved, and state is installed in network devices.

Signaling Protocols:

Connection-oriented networks require signaling protocols to establish and tear down connections. These operate on a control plane separate from the data plane.

Example Signaling Flow (Generic Virtual Circuit):

vc_signaling.pseudo
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
// Connection Establishment - Source Initiation
function establish_virtual_circuit(destination, qos_requirements):
    // Step 1: Source creates SETUP message
    setup_msg = create_setup_message(
        source_address = my_address,
        destination_address = destination,
        requested_qos = qos_requirements,
        call_reference = generate_unique_id()
    )
    
    // Step 2: Send to first switch
    next_hop = routing_table.lookup(destination)
    send(setup_msg, next_hop)
    
    // Step 3: Wait for response
    response = wait_for_response(timeout = 30 seconds)
    
    if response.type == CONNECT:
        // Connection established - extract assigned VCI
        my_vci = response.vci
        connection_state = ESTABLISHED
        return Connection(vci = my_vci, destination = destination)
    
    else if response.type == RELEASE_COMPLETE:
        // Connection rejected
        error_cause = response.cause
        raise ConnectionRefused(error_cause)
 
// At Each Intermediate Switch
function process_setup_message(setup_msg, incoming_port):
    destination = setup_msg.destination_address
    qos = setup_msg.requested_qos
    
    // Check if we can meet QoS requirements
    if not can_satisfy_qos(qos):
        send_release(incoming_port, cause="Insufficient resources")
        return
    
    // Allocate local resources
    incoming_vci = allocate_vci(incoming_port)
    outgoing_port = routing_table.lookup(destination)
    outgoing_vci = allocate_vci(outgoing_port)
    
    // Install forwarding table entry
    forwarding_table.add_entry(
        incoming_port, incoming_vci,
        outgoing_port, outgoing_vci
    )
    
    // Reserve QoS resources
    reserve_bandwidth(outgoing_port, qos.bandwidth)
    allocate_buffers(qos.buffer_requirement)
    
    // Forward setup toward destination
    setup_msg.vci = outgoing_vci
    send(setup_msg, outgoing_port)

Resource Reservation:

One key advantage of connection-oriented service is the ability to reserve resources during connection setup.

What Can Be Reserved:

Bandwidth: Guaranteed data rate on each link
Buffer Space: Queue capacity for this connection's packets
Processing Capacity: CPU time for packet handling
Latency Bounds: Maximum delay through each switch

Call Admission Control (CAC):

When a new connection request arrives, switches perform Call Admission Control:

Determine if sufficient resources exist to meet QoS requirements
If yes: reserve resources and forward the setup
If no: reject the connection request

This ensures that accepted connections can meet their performance guarantees.

The Setup Delay Tradeoff

Connection establishment takes time—signaling messages must traverse the network, and each switch must process them. This 'call setup delay' might be 100ms-1s depending on path length. For long-lived connections (hours-long calls), this is negligible. For short transactions (web page fetch), it can be problematic.

Advantages of Connection-Oriented Service

Connection-oriented service provides capabilities that are difficult or impossible to achieve with connectionless approaches.

Guaranteed Quality of Service

•Hard Bandwidth Guarantees — Reserved capacity cannot be preempted by other traffic. The connection gets its promised rate.
•Bounded Latency — With reserved resources and fixed paths, maximum delay is predictable and guaranteed.
•Low Jitter — Same path + reserved resources = consistent delay, essential for real-time media.
•No Packet Loss from Congestion — If admission control works correctly, accepted connections never experience congestion loss.
•SLA Enforcement — Service Level Agreements can be precisely defined and guaranteed.

Efficient Forwarding

•Short Labels — VCIs/labels are much shorter than addresses, encoded faster lookup.
•Array-Based Lookup — Labels can be direct table indices → O(1) lookup vs. longest-prefix matching.
•Simplified Header — Data packets don't need full addressing; just the label.
•Hardware-Friendly — Fixed-size labels enable efficient hardware implementation.

Traffic Engineering

•Explicit Path Control — Network operators can specify exactly which path traffic takes.
•Load Balancing — Traffic can be distributed across multiple paths deliberately.
•Failure Protection — Backup paths can be pre-established for fast failover.
•Capacity Planning — Known paths enable accurate capacity modeling.

When Connection-Oriented Shines

Connection-oriented service excels for: (1) Real-time applications requiring guaranteed performance (voice, video, industrial control), (2) Premium services where SLAs must be mathematically guaranteed, (3) Carrier networks requiring traffic engineering, (4) Long-duration communications where setup overhead is amortized.

Disadvantages and Challenges

Despite its advantages, connection-oriented service has significant drawbacks that explain why it didn't become the universal network layer model.

State Requirements

•Per-Connection State in Switches — Each active connection requires a forwarding table entry. With millions of connections, this becomes substantial memory.
•State Synchronization — If a switch crashes, it loses connection state. Recovery requires re-establishing or re-synchronizing connections.
•State Cleanup — If endpoints crash without proper teardown, 'orphan' connection state remains until timeouts.
•Scaling Limits — Maximum simultaneous connections limited by switch table sizes.

Failure Handling

•Single Point of Failure — If any switch on the path fails, the entire connection breaks. Rerouting requires re-establishing the connection.
•Slow Recovery — Failure recovery requires signaling to set up new paths—slower than connectionless automatic rerouting.
•Path Rigidity — Traffic cannot flow around congestion; it's locked to the established path.
•Complex Failover — Fast protection switching requires pre-establishing backup paths, doubling state.

Overhead and Complexity

•Setup Delay — Short-lived communications suffer disproportionate setup overhead.
•Signaling Complexity — Signaling protocols add complexity to switches and the overall system.
•Resource Waste — Reserved but unused bandwidth is wasted—can't be used by other traffic.
•All-or-Nothing Admission — Either the full QoS can be met, or the connection is rejected—no graceful degradation.

Why the Internet Chose Connectionless

The Internet's designers explicitly rejected connection-oriented networking because it requires homogeneous networks (all switches must support the same signaling), limits resilience (path failures break connections), and scales poorly with billions of simultaneous communications. The stateless datagram model proved more suitable for a heterogeneous, resilient, global network.

Summary: Connection-Oriented Service

Connection-oriented service represents an alternative philosophy to connectionless datagram networks. Let's consolidate the essential concepts:

Key Takeaways

•Three-Phase Lifecycle — Connection-oriented service has distinct establishment, data transfer, and termination phases.
•Virtual Circuits — A VC is a logical path through the network with per-connection state in each switch.
•Label Switching — VCIs/labels enable fast, simple forwarding without full address processing.
•SVCs vs. PVCs — Switched VCs are on-demand; Permanent VCs are pre-configured for stable traffic.
•Resource Reservation — Setup phase can reserve bandwidth, buffers, and guarantee QoS.
•Stateful Networking — Switches maintain per-connection state, limiting scale but enabling guarantees.
•Tradeoffs — Guarantees come at the cost of flexibility, resilience, and state management complexity.

What's Next:

With both connectionless and connection-oriented service models understood, we'll examine what the network layer actually guarantees: best-effort delivery. Understanding best-effort service—what it means, what it provides, and what it explicitly doesn't—is essential for building reliable systems on unreliable foundations.

Page Complete

You now understand connection-oriented service at the network layer. You can explain virtual circuits, signaling, label switching, and articulate when connection-oriented approaches are appropriate versus connectionless alternatives. This knowledge is crucial for understanding MPLS, carrier networks, and QoS architectures.

2 / 5

Loading learning content...

Computer NetworksNetwork Layer Services

Network Layer Services

LevelIntermediate

Duration75 mins

TopicNetwork Layer Services

2 / 5

Connection-Oriented Service

Establishing a Path Before Communication

What You Will Learn

Defining Connection-Oriented Service

The Core Definition:

A connection-oriented network service operates through three distinct phases:

Connection Establishment — A signaling process sets up the path through the network. Intermediate switches/routers are configured with forwarding information for this specific connection.
Data Transfer — Data flows along the established path. All packets follow the same route in sequence.
Connection Termination — When communication ends, a teardown process releases resources and removes connection state from network devices.

Key Characteristics:

Fundamental Properties

•Explicit Setup Phase — Before data transfer, signaling messages traverse the network to establish the connection and configure intermediate devices.
•Stateful Network Devices — Switches/routers maintain per-connection state: connection identifiers, forwarding entries, possibly reserved resources.
•Fixed Path — All packets of a connection follow the same path, ensuring in-order delivery (barring errors).
•Connection Identifier — Instead of full destination addresses, packets carry short connection identifiers (labels) that switches use for fast forwarding.
•Guaranteed Ordering — Because packets follow the same path, they arrive in the order sent (unless lost).
•Possible Resource Reservation — The connection setup can reserve bandwidth, buffer space, or processing capacity along the path.

Virtual Circuits vs. Physical Circuits

The Virtual Circuit Model

A virtual circuit (VC) is the fundamental abstraction of connection-oriented networking. It represents a logical path through the network that behaves like a dedicated connection.

How Virtual Circuits Work:

1. Connection Request: The source sends a setup message containing the destination address. This message travels through the network, with each switch:

Allocating local resources (buffer space, table entries)
Assigning a Virtual Circuit Identifier (VCI) for this connection on each link
Adding an entry to its forwarding table: (incoming port, incoming VCI) → (outgoing port, outgoing VCI)
Forwarding the setup message toward the destination

2. Connection Confirmation: When the setup reaches the destination, a confirmation travels back, finalizing the path. Now all switches along the route have forwarding entries for this VC.

3. Data Transfer: Packets carry only the VCI (not full addresses). Each switch:

Looks up (incoming port, VCI) in its table
Finds (outgoing port, new VCI)
Replaces the VCI with the new value
Forwards the packet

4. Teardown: When communication ends, teardown messages traverse the path, and switches delete their VC entries.

Converting Mermaid diagram...

Virtual Circuit Identifier (VCI):

Why local labels?

Labels can be short (typically 16-32 bits vs. 32+ for addresses)
Fast table lookup (labels can be array indices)
Same label can be reused on different links for different VCs

Example Switch Forwarding Table (Virtual Circuit)
Incoming Port	Incoming VCI	Outgoing Port	Outgoing VCI
Port 0	12	Port 2	41
Port 0	25	Port 1	18
Port 1	33	Port 2	55
Port 2	87	Port 0	12

Types of Virtual Circuits

Virtual circuits come in two primary forms, each suited to different usage patterns.

Switched Virtual Circuits (SVCs):

SVCs are dynamically established on demand, like phone calls. When communication is needed, the source initiates connection setup. When finished, the connection is torn down.

Characteristics:

Established per-session
Setup and teardown overhead for each connection
Resources may be allocated only for the connection duration
Flexible—no pre-configuration needed
Used for sporadic, on-demand communication

Permanent Virtual Circuits (PVCs):

PVCs are pre-configured connections that remain established indefinitely. They're set up administratively (not by signaling) and stay active continuously.

Characteristics:

Configured by network administrators
No per-session setup delay
Resources permanently allocated
Lower flexibility—changes require reconfiguration
Ideal for constant communication between fixed endpoints
Common in enterprise WAN connections (e.g., between branch offices)

Switched Virtual Circuits

•Established dynamically via signaling
•Setup time before data can flow
•Resources allocated per-session
•Automatic teardown when done
•Scales well for many destinations
•Phone calls, video conferences

Permanent Virtual Circuits

•Pre-configured administratively
•No setup delay—always ready
•Resources permanently reserved
•Changes require manual config
•Efficient for fixed endpoints
•Branch office connections

Hybrid Approaches

Technologies Using Connection-Oriented Service

Several major networking technologies employ connection-oriented service principles. Understanding these examples illustrates where and why this model is used.

ATM (Asynchronous Transfer Mode):

ATM was a comprehensive connection-oriented technology designed in the 1990s to unify voice, video, and data networking.

Key Features:

Fixed-size 53-byte cells (48 payload + 5 header)
Virtual Paths (VP) and Virtual Channels (VC)
Hardware-based cell switching
Built-in QoS with multiple service categories
Signaling protocol for SVC establishment

ATM saw widespread deployment in carrier networks but lost to IP-based solutions for end-to-end communication due to complexity and overhead.

MPLS (Multiprotocol Label Switching):

MPLS takes the best of both worlds—connectionless IP for universal addressing with label-switching for efficient forwarding and traffic engineering.

Key Features:

Labels inserted between Layer 2 and Layer 3 headers
Label switching provides connection-like forwarding
Label Distribution Protocol (LDP) or RSVP-TE for path setup
Traffic engineering enables explicit path control
Widely deployed in carrier and enterprise networks

MPLS is sometimes called 'Layer 2.5' because it operates between traditional Layer 2 and Layer 3.

mpls_label_structure.txt
MPLS Label Stack Entry (32 bits):
+---------------+---+---+--------+
|     Label     |TC |S  |  TTL   |
+---------------+---+---+--------+
|   20 bits     |3b |1b | 8 bits |
 
Label: Virtual circuit identifier (0-1,048,575)
TC: Traffic Class (QoS marking)
S: Bottom of Stack flag (1 = last label)
TTL: Time To Live (prevents loops)
 
MPLS Packet Structure:
+==============+==============+==============+
| Layer 2 Hdr  | MPLS Labels  | IP Packet    |
| (Ethernet)   | (1 or more)  | (original)   |
+==============+==============+==============+

Frame Relay:

Frame Relay was a connection-oriented WAN technology popular in the 1990s and 2000s.

Key Features:

Data Link Connection Identifier (DLCI) for virtual circuits
Variable-length frames (more efficient than ATM cells)
Primarily PVCs (SVCs were defined but rarely used)
Committed Information Rate (CIR) for bandwidth guarantees
Largely replaced by MPLS and Ethernet services

Traditional Telephony (PSTN):

The Public Switched Telephone Network exemplifies connection-oriented design:

Call setup establishes a circuit
64 kbps reserved per call direction
All voice samples follow the same path
Call teardown releases resources

Connection Establishment in Detail

The connection establishment phase is critical—it's where paths are computed, resources are reserved, and state is installed in network devices.

Signaling Protocols:

Connection-oriented networks require signaling protocols to establish and tear down connections. These operate on a control plane separate from the data plane.

Example Signaling Flow (Generic Virtual Circuit):

vc_signaling.pseudo
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
// Connection Establishment - Source Initiation
function establish_virtual_circuit(destination, qos_requirements):
    // Step 1: Source creates SETUP message
    setup_msg = create_setup_message(
        source_address = my_address,
        destination_address = destination,
        requested_qos = qos_requirements,
        call_reference = generate_unique_id()
    )
    
    // Step 2: Send to first switch
    next_hop = routing_table.lookup(destination)
    send(setup_msg, next_hop)
    
    // Step 3: Wait for response
    response = wait_for_response(timeout = 30 seconds)
    
    if response.type == CONNECT:
        // Connection established - extract assigned VCI
        my_vci = response.vci
        connection_state = ESTABLISHED
        return Connection(vci = my_vci, destination = destination)
    
    else if response.type == RELEASE_COMPLETE:
        // Connection rejected
        error_cause = response.cause
        raise ConnectionRefused(error_cause)
 
// At Each Intermediate Switch
function process_setup_message(setup_msg, incoming_port):
    destination = setup_msg.destination_address
    qos = setup_msg.requested_qos
    
    // Check if we can meet QoS requirements
    if not can_satisfy_qos(qos):
        send_release(incoming_port, cause="Insufficient resources")
        return
    
    // Allocate local resources
    incoming_vci = allocate_vci(incoming_port)
    outgoing_port = routing_table.lookup(destination)
    outgoing_vci = allocate_vci(outgoing_port)
    
    // Install forwarding table entry
    forwarding_table.add_entry(
        incoming_port, incoming_vci,
        outgoing_port, outgoing_vci
    )
    
    // Reserve QoS resources
    reserve_bandwidth(outgoing_port, qos.bandwidth)
    allocate_buffers(qos.buffer_requirement)
    
    // Forward setup toward destination
    setup_msg.vci = outgoing_vci
    send(setup_msg, outgoing_port)

Resource Reservation:

One key advantage of connection-oriented service is the ability to reserve resources during connection setup.

What Can Be Reserved:

Bandwidth: Guaranteed data rate on each link
Buffer Space: Queue capacity for this connection's packets
Processing Capacity: CPU time for packet handling
Latency Bounds: Maximum delay through each switch

Call Admission Control (CAC):

When a new connection request arrives, switches perform Call Admission Control:

Determine if sufficient resources exist to meet QoS requirements
If yes: reserve resources and forward the setup
If no: reject the connection request

This ensures that accepted connections can meet their performance guarantees.

The Setup Delay Tradeoff

Advantages of Connection-Oriented Service

Connection-oriented service provides capabilities that are difficult or impossible to achieve with connectionless approaches.

Guaranteed Quality of Service

•Hard Bandwidth Guarantees — Reserved capacity cannot be preempted by other traffic. The connection gets its promised rate.
•Bounded Latency — With reserved resources and fixed paths, maximum delay is predictable and guaranteed.
•Low Jitter — Same path + reserved resources = consistent delay, essential for real-time media.
•No Packet Loss from Congestion — If admission control works correctly, accepted connections never experience congestion loss.
•SLA Enforcement — Service Level Agreements can be precisely defined and guaranteed.

Efficient Forwarding

•Short Labels — VCIs/labels are much shorter than addresses, encoded faster lookup.
•Array-Based Lookup — Labels can be direct table indices → O(1) lookup vs. longest-prefix matching.
•Simplified Header — Data packets don't need full addressing; just the label.
•Hardware-Friendly — Fixed-size labels enable efficient hardware implementation.

Traffic Engineering

•Explicit Path Control — Network operators can specify exactly which path traffic takes.
•Load Balancing — Traffic can be distributed across multiple paths deliberately.
•Failure Protection — Backup paths can be pre-established for fast failover.
•Capacity Planning — Known paths enable accurate capacity modeling.

When Connection-Oriented Shines

Disadvantages and Challenges

Despite its advantages, connection-oriented service has significant drawbacks that explain why it didn't become the universal network layer model.

State Requirements

•Per-Connection State in Switches — Each active connection requires a forwarding table entry. With millions of connections, this becomes substantial memory.
•State Synchronization — If a switch crashes, it loses connection state. Recovery requires re-establishing or re-synchronizing connections.
•State Cleanup — If endpoints crash without proper teardown, 'orphan' connection state remains until timeouts.
•Scaling Limits — Maximum simultaneous connections limited by switch table sizes.

Failure Handling

•Single Point of Failure — If any switch on the path fails, the entire connection breaks. Rerouting requires re-establishing the connection.
•Slow Recovery — Failure recovery requires signaling to set up new paths—slower than connectionless automatic rerouting.
•Path Rigidity — Traffic cannot flow around congestion; it's locked to the established path.
•Complex Failover — Fast protection switching requires pre-establishing backup paths, doubling state.

Overhead and Complexity

•Setup Delay — Short-lived communications suffer disproportionate setup overhead.
•Signaling Complexity — Signaling protocols add complexity to switches and the overall system.
•Resource Waste — Reserved but unused bandwidth is wasted—can't be used by other traffic.
•All-or-Nothing Admission — Either the full QoS can be met, or the connection is rejected—no graceful degradation.

Why the Internet Chose Connectionless

Summary: Connection-Oriented Service

Connection-oriented service represents an alternative philosophy to connectionless datagram networks. Let's consolidate the essential concepts:

Key Takeaways

•Three-Phase Lifecycle — Connection-oriented service has distinct establishment, data transfer, and termination phases.
•Virtual Circuits — A VC is a logical path through the network with per-connection state in each switch.
•Label Switching — VCIs/labels enable fast, simple forwarding without full address processing.
•SVCs vs. PVCs — Switched VCs are on-demand; Permanent VCs are pre-configured for stable traffic.
•Resource Reservation — Setup phase can reserve bandwidth, buffers, and guarantee QoS.
•Stateful Networking — Switches maintain per-connection state, limiting scale but enabling guarantees.
•Tradeoffs — Guarantees come at the cost of flexibility, resilience, and state management complexity.

What's Next:

Page Complete

2 / 5