Packet Delivery - Learning Module

Loading content...

0/228

Unicast Delivery

The Foundation of Network Communication

Every time you send an email, visit a website, or make a video call, your data embarks on a journey through the network layer. But how does this layer ensure your packet reaches its intended destination? The answer lies in understanding packet delivery types—the fundamental mechanisms that determine how data travels from source to destination.

In the vast landscape of network communication, unicast delivery stands as the cornerstone—the most prevalent and fundamental form of packet transmission. It represents the simple yet powerful concept of one-to-one communication: a single source sending data to a single destination.

This page will take you on a deep exploration of unicast delivery, examining not just what it is, but how it works at every level—from addressing and routing to implementation considerations and real-world applications.

What You Will Learn

By the end of this page, you will understand: (1) The precise definition and characteristics of unicast delivery, (2) How unicast addresses are structured and interpreted, (3) The routing mechanisms that enable unicast packet forwarding, (4) When and why to use unicast over other delivery types, and (5) The performance implications and optimization strategies for unicast communication.

Defining Unicast Delivery

Unicast delivery is the process of sending a packet from exactly one source host to exactly one destination host. The term 'unicast' combines the Latin prefix 'uni-' (meaning 'one') with 'cast' (meaning 'to throw or send'), perfectly capturing the one-to-one nature of this communication pattern.

At the network layer, unicast represents the default and most common delivery mode. When you specify a destination IP address that identifies a single specific interface, you're engaging in unicast communication. The network infrastructure then assumes the responsibility of forwarding your packet through the appropriate path to reach that single destination.

The Essence of Unicast

Unicast is characterized by three fundamental properties: (1) Single Source — exactly one sender initiates the communication, (2) Single Destination — exactly one receiver is targeted, and (3) Unique Identification — the destination is identified by a unique address that maps to a single network interface.

The Communication Model

The unicast communication model follows a straightforward pattern:

Address Resolution: The source determines the network layer address (IP address) of the intended destination
Packet Construction: The source encapsulates data with source and destination IP addresses in the packet header
Routing Decision: Each router along the path examines the destination address and forwards accordingly
Delivery: The packet arrives at the destination host, where the network layer delivers it to the appropriate upper layer

This model may seem simple, but its elegance lies in its scalability. The same mechanism works whether the destination is on the same local network or across the globe, separated by hundreds of routers.

Unicast Delivery Characteristics
Characteristic	Description	Implication
Source Count	Exactly one sender	Clear origin for responses and troubleshooting
Destination Count	Exactly one receiver	Deterministic delivery target
Address Type	Host-specific identifier	Globally unique destination identification
Routing Behavior	Standard hop-by-hop forwarding	Utilizes full routing infrastructure
Scalability	Linear relationship: one packet per destination	Efficient for single-recipient communication
Reliability	Can be made reliable (TCP) or unreliable (UDP)	Flexible transport layer options

Unicast Addressing in Depth

Understanding unicast delivery requires a deep dive into how unicast addresses are structured and interpreted. In IP networks, unicast addresses form the vast majority of the address space and follow specific structural rules that enable efficient routing.

IPv4 Unicast Address Space

In IPv4, unicast addresses occupy most of the 32-bit address space. Excluding reserved ranges (multicast, broadcast, loopback, private, etc.), unicast addresses are routable addresses that identify a single specific interface on the global Internet or a private network.

The hierarchical structure of IP addresses is fundamental to understanding unicast operation:

IPv4 Address Structure
IPv4 Address: 192.168.1.100 (32 bits total)
┌──────────────────────────────────────────────────┐
│        Network Portion        │   Host Portion   │
│       (Identifies network)    │ (Identifies host)│
└──────────────────────────────────────────────────┘
 
Example with /24 subnet mask (255.255.255.0):
 
    192    .    168    .    1      .    100
  ┌─────────────────────────────┬───────────────┐
  │     Network ID: 192.168.1   │   Host ID: 100│
  │     (24 bits)               │   (8 bits)    │
  └─────────────────────────────┴───────────────┘
 
Routing Decision:
  - Routers examine the NETWORK portion to forward packets
  - The HOST portion identifies the specific interface within that network
  
Key Insight:
  Unicast Address = Network Prefix + Host Identifier
  This hierarchy enables scalable routing - routers don't need
  to know about every host, just every network prefix.

IPv6 Unicast Addressing

IPv6 introduces a more sophisticated unicast addressing model with 128-bit addresses. The vastly larger address space allows for more flexible allocation and built-in support for different unicast scopes:

Global Unicast Addresses (GUA): Routable across the entire Internet, equivalent to public IPv4 addresses
- Prefix: 2000::/3 (addresses starting with 2 or 3)
- Structure: 48-bit global routing prefix + 16-bit subnet ID + 64-bit interface ID
Link-Local Addresses: Valid only within a single link (network segment)
- Prefix: FE80::/10
- Automatically generated on every IPv6-enabled interface
- Essential for neighbor discovery and local communication
Unique Local Addresses (ULA): Private addresses for internal use
- Prefix: FC00::/7 (typically FD00::/8 in practice)
- Analogous to IPv4 private address ranges

Interface Identifier Generation

In IPv6, the 64-bit interface identifier can be generated using several methods: (1) Modified EUI-64 — derived from the MAC address with a FFFE insertion and bit flip, (2) Random — cryptographically random for privacy (RFC 4941), (3) Static — manually configured by administrators. This flexibility addresses both operational needs and privacy concerns.

IPv4 vs IPv6 Unicast Address Comparison
Aspect	IPv4 Unicast	IPv6 Unicast
Address Size	32 bits (4.3 billion addresses)	128 bits (3.4×10³⁸ addresses)
Notation	Dotted decimal (192.168.1.1)	Colon hexadecimal (2001:db8::1)
Auto-configuration	DHCP required	SLAAC built-in (stateless)
Private Ranges	10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16	FC00::/7 (Unique Local)
Link-Local	169.254.0.0/16 (APIPA)	FE80::/10 (mandatory)
Global Unicast	Public routable addresses	2000::/3

Routing Mechanism for Unicast

The journey of a unicast packet from source to destination is orchestrated by the routing mechanism—a distributed decision-making process that determines the optimal path through the network. Understanding this mechanism is crucial for grasping how unicast delivery actually works in practice.

The Forwarding Decision

When a router receives a unicast packet, it must make a forwarding decision: which interface should this packet be sent out on to reach its destination? This decision follows a precise algorithm:

Unicast Forwarding Algorithm
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
UNICAST_FORWARD(packet):
    destination_ip = packet.header.destination_address
    
    # Step 1: Check local delivery
    IF destination_ip matches any local interface:
        DELIVER to upper layer protocol
        RETURN
    
    # Step 2: Consult routing table
    best_match = NULL
    best_prefix_length = -1
    
    FOR each entry in routing_table:
        IF destination_ip matches entry.network_prefix:
            IF entry.prefix_length > best_prefix_length:
                best_match = entry
                best_prefix_length = entry.prefix_length
    
    # Step 3: Forward or drop
    IF best_match != NULL:
        next_hop = best_match.next_hop_address
        output_interface = best_match.output_interface
        
        # Decrement TTL
        packet.header.TTL = packet.header.TTL - 1
        IF packet.header.TTL == 0:
            SEND ICMP Time Exceeded to source
            DROP packet
            RETURN
        
        # Recalculate header checksum
        packet.header.checksum = recalculate_checksum()
        
        # Resolve next-hop MAC address (ARP/NDP)
        next_hop_mac = resolve_layer2_address(next_hop)
        
        # Transmit on output interface
        SEND packet via output_interface to next_hop_mac
    ELSE:
        # No route to destination
        SEND ICMP Destination Unreachable to source
        DROP packet

Longest Prefix Match (LPM)

The longest prefix match algorithm is the cornerstone of IP routing. When multiple routing table entries could match a destination address, the router selects the entry with the longest (most specific) prefix. This enables hierarchical routing and route aggregation.

Why Longest Prefix Match?

Consider a router with entries for 10.0.0.0/8 (via Gateway A) and 10.1.0.0/16 (via Gateway B). A packet destined for 10.1.2.3 matches BOTH entries. LPM selects 10.1.0.0/16 because its 16-bit prefix is longer (more specific) than the 8-bit prefix. This allows exceptions and more specific routes to override general routes.

Converting Mermaid diagram...

The Complete Unicast Path

A unicast packet traverses the network through a series of hops. Each hop represents a router that makes an independent forwarding decision. This distributed approach provides:

Resilience: If one path fails, routing protocols can find alternatives
Scalability: No single device needs complete network knowledge
Adaptability: Routes can change dynamically based on network conditions

However, this hop-by-hop forwarding also means that each router independently decides the next hop. The source has limited control over the path (except via source routing, which is rarely used for security reasons).

Unicast in the IP Header

Every unicast packet carries essential information in its IP header that enables proper delivery. Understanding which header fields are critical for unicast operation provides insight into how the network layer achieves its goals.

Critical Header Fields for Unicast

IP Header Fields Relevant to Unicast Delivery
Field	Size (IPv4)	Role in Unicast
Source Address	32 bits	Identifies the originating host; used for replies and error messages
Destination Address	32 bits	The unicast target; drives all routing decisions
TTL/Hop Limit	8 bits	Prevents infinite loops; decremented at each hop
Protocol/Next Header	8 bits	Identifies upper-layer protocol (TCP, UDP, ICMP)
Header Checksum	16 bits (IPv4 only)	Ensures header integrity; recalculated at each hop due to TTL change
Identification	16 bits	Groups fragments of the same original packet
Fragment Offset	13 bits	Position of this fragment in original datagram

IPv4 Header Structure
 0                   1                   2                   3
 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|Version|  IHL  |Type of Service|          Total Length         |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|         Identification        |Flags|      Fragment Offset    |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|  Time to Live |    Protocol   |         Header Checksum       |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                       Source Address                          | ◄── Sender's 
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+    unicast addr
|                    Destination Address                        | ◄── Target's
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+    unicast addr
|                    Options (if IHL > 5)                       |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
For UNICAST packets:
• Destination Address = specific host interface (not broadcast/multicast)
• Source Address = sender's interface address
• Both addresses are globally or locally unique unicast addresses

Source Address Validation

Networks implement ingress filtering to verify that the source address in unicast packets is legitimate. Packets with spoofed source addresses can be used for attacks (reflection, amplification). BCP 38/RFC 2827 recommends that ISPs filter packets with source addresses that don't belong to their address space.

When to Use Unicast

While unicast is the default and most common delivery type, understanding when it's the optimal choice versus other delivery types is essential for network design and application development.

Unicast is Ideal When:

Unicast Use Cases

•Single Recipient Communication — The most obvious case: email, web browsing, file downloads, or any scenario where data is intended for one specific host.
•Bidirectional Communication — Protocols requiring request-response patterns (HTTP, SSH, FTP) naturally use unicast because each party needs to address the other specifically.
•Reliable Delivery Required — When using TCP for guaranteed delivery, unicast is mandatory. TCP's connection-oriented design requires exactly two endpoints.
•Security-Sensitive Data — Private or encrypted communications where data should reach only one recipient, such as banking transactions or personal messages.
•Different Content Per Recipient — When each recipient needs personalized or unique content, unicast is necessary (e.g., user-specific dashboard data).

Unicast Advantages

•Simple addressing model
•Well-supported by all networks
•Reliable delivery possible (TCP)
•Easy troubleshooting and tracing
•Full routing protocol support
•End-to-end security with TLS

Unicast Limitations

•Inefficient for one-to-many
•Sender bandwidth scales with recipients
•No native service redundancy
•Every recipient needs separate packet
•High source load for popular content
•CDN/caching needed for scale

The Scalability Trade-off

If you need to send identical content to 1,000 recipients using unicast, you must send 1,000 copies of the data—consuming 1,000× the bandwidth at the source. This is why content delivery networks (CDNs) and multicast exist: to avoid this linear scaling problem. However, for personalized or bidirectional communication, unicast remains the only practical choice.

Real-World Applications

Unicast delivery underpins the vast majority of Internet traffic. Let's examine how it manifests in real-world applications and the considerations that make it work at scale.

Unicast in Production Systems

•Web Traffic (HTTP/HTTPS) — Every web page request and response uses unicast. Your browser establishes a TCP connection to a web server's IP address, and all data flows via unicast packets. Even when accessing a CDN, the final hop is unicast to your device.
•Email Delivery (SMTP/IMAP/POP3) — Email traverses the Internet via unicast connections between mail servers, and from mail servers to clients. Each recipient's MX record resolves to a unicast address.
•Video Streaming (Netflix, YouTube) — Though you might expect multicast for video, most streaming uses unicast with adaptive bitrate (ABR). Each viewer receives a personalized stream based on their bandwidth, device, and viewing position.
•VoIP and Video Calls — Real-time communication relies on unicast, often over UDP for low latency. WebRTC, Zoom, and similar platforms establish peer-to-peer or client-server unicast flows.
•Database Queries — Application servers communicate with databases via unicast TCP connections. Each query and response is a unicast exchange between specific endpoints.

Unicast Traffic Distribution by Application
Application	Protocol	Unicast Role	Scale Considerations
Web Browsing	HTTPS/TCP	Request every resource	CDN edge servers reduce latency
Streaming Video	HTTPS/QUIC	Per-viewer adaptive stream	ABR adjusts quality per user
Online Gaming	UDP/TCP	Game state synchronization	Low latency prioritized
Cloud Services	Various/TCP	API calls, data transfer	Load balancers distribute connections
Remote Work (VPN)	IPsec/SSL	Encrypted tunnel	Split tunneling optimizes traffic

The CDN Pattern

Content Delivery Networks (CDNs) solve the unicast scaling problem not by replacing unicast, but by distributing the unicast endpoints. Instead of one origin server sending unicast to millions of users, hundreds of edge servers each handle unicast to regional users. The content is replicated to edges, but delivery remains unicast. This is why services like Netflix can serve billions of hours of video using 'just' unicast.

Performance Considerations

Optimizing unicast performance involves understanding the factors that affect packet delivery speed, reliability, and efficiency. Let's examine the key considerations that network engineers and application developers must address.

Latency Components

The total time for a unicast packet to travel from source to destination consists of several components:

Unicast Latency Breakdown
Total Latency = Σ(at each hop and link):
                 
    ┌─────────────────────────────────────────────────────────────┐
    │ PROCESSING DELAY                                            │
    │ • Time to examine packet header                             │
    │ • Routing table lookup (typically microseconds)             │
    │ • Modern hardware: ~1-10 μs per hop                        │
    ├─────────────────────────────────────────────────────────────┤
    │ QUEUING DELAY                                               │
    │ • Time waiting in output queue                              │
    │ • Highly variable (depends on congestion)                   │
    │ • Can range from 0 to hundreds of milliseconds              │
    ├─────────────────────────────────────────────────────────────┤
    │ TRANSMISSION DELAY                                          │
    │ • Time to push bits onto the link                          │
    │ • = Packet Size / Link Bandwidth                           │
    │ • 1500 bytes @ 1 Gbps = 12 μs                              │
    ├─────────────────────────────────────────────────────────────┤
    │ PROPAGATION DELAY                                           │
    │ • Time for signal to travel physical distance               │
    │ • ≈ Distance / Speed of Light in medium                    │
    │ • ~5 μs per km in fiber                                    │
    └─────────────────────────────────────────────────────────────┘
 
Example: London to Sydney (~17,000 km via undersea cable)
    Propagation alone: 17,000 km × 5 μs/km ≈ 85 ms (one way)
    Round-trip minimum: ~170 ms (physics limit)

Optimization Strategies

Network engineers employ several strategies to optimize unicast delivery:

Unicast Optimization Techniques

•Route Optimization — Using BGP policies to select lower-latency paths, even if longer in hop count. Some providers offer 'latency-based routing' for premium traffic.
•Traffic Engineering — MPLS and segment routing allow explicit path control, avoiding congested links and ensuring predictable performance.
•Edge Computing — Moving services closer to users reduces propagation delay. A CDN edge server 50 km away beats an origin server 5000 km away.
•Connection Reuse — HTTP/2, HTTP/3, and connection pooling amortize connection setup costs across multiple requests.
•Quality of Service (QoS) — Prioritizing latency-sensitive unicast traffic (VoIP, gaming) over bulk transfers reduces queuing delay.

The Speed of Light Limit

No optimization can overcome physics. Light in fiber travels at about 200,000 km/s (two-thirds the vacuum speed of light). The London-Sydney example shows that even with perfect hardware, a round trip takes ~170 ms. For applications requiring <50 ms latency, servers MUST be geographically close to users. This is why global applications deploy in multiple regions.

Summary: Unicast Delivery

We've explored unicast delivery in comprehensive detail. Let's consolidate the essential knowledge:

Key Takeaways

•Unicast is one-to-one communication — A single source addresses a single destination using a globally or locally unique IP address.
•Addressing drives routing — The destination IP address in the packet header determines how routers forward the packet hop-by-hop.
•Longest Prefix Match — Routers select the most specific matching route, enabling hierarchical routing and efficient forwarding tables.
•IPv4 and IPv6 differences — IPv6 introduces mandatory link-local addresses, larger global unicast space, and built-in auto-configuration.
•Scalability is linear — Unicast requires one packet per recipient, which is efficient for single-recipient communication but problematic for mass distribution.
•Latency has physical limits — Propagation delay imposes minimum latency based on distance; geographic distribution is essential for low-latency applications.
•Unicast dominates Internet traffic — Despite alternatives, most applications rely on unicast due to its simplicity, reliability options, and universal support.

What's next:

Unicast addresses a single host, but many network scenarios require reaching all hosts on a network simultaneously. The next page explores broadcast delivery—the one-to-all communication pattern that, while limited in scope, plays a critical role in network operations like address resolution and service discovery.

Page Complete

You now have a deep understanding of unicast delivery—from addressing fundamentals to routing mechanisms to real-world performance considerations. This foundation prepares you to contrast unicast with broadcast, multicast, and anycast delivery types in the upcoming pages.