Computer Networks (CN) — CIE-3 Complete 30/30 Exam Guide
Exam Scope Overview:
- Module 3: Starts from Transport Layer (Transport Service, Elements of Transport Protocols, Connection Establishment & Release, Multiplexing & Demultiplexing, Crash Recovery, UDP, TCP Segment Structure, RTT Estimation & Timeout, Congestion Control) + IPv4 vs IPv6 Differences.
- Module 4: Application Layer (HTTP Persistent vs Non-Persistent, Conditional GET, Cookies, Web Caching, FTP, SMTP vs HTTP, POP3/IMAP, DNS) + AI in Computer Networks (AI in SDN & Self-Healing Networks).
Master Checklist of Important CN Questions
Module 3:
- RTT Estimation & TCP Timeout Numerical (also HTTP RTT calculation)
- TCP Segment Structure (neat diagram + explain every field)
- TCP Three-Way Handshake (connection establishment + handling old duplicates)
- Transport-Layer Multiplexing & Demultiplexing (UDP 2-tuple vs TCP 4-tuple)
- Difference between IPv4 and IPv6
- Crash Recovery (ACK-first vs Write-first & Forbidden Region)
Module 4:
- HTTP: Persistent vs Non-Persistent Connections (+ 2M Quick Hits: Stateless, Conditional GET, Status Codes)
- Cookies — How Information Is Stored & User-Server Interaction
- DNS — Function, Types/Classes of DNS Servers & Resource Records
- Difference between HTTP and SMTP
- SDN & AI in SDN
- Self-Healing Networks
Module 3 — High-Priority Questions & Exam Answers
1. RTT Estimation & TCP Timeout Problem (Guaranteed Numerical)
Formulas (RFC 6298):
-
EstimatedRTT (Exponential Weighted Moving Average):
EstimatedRTT = (1 - α) × EstimatedRTT + α × SampleRTT(recommended α = 0.125 = 1/8) -
DevRTT (RTT variability / deviation):
DevRTT = (1 - β) × DevRTT + β × |SampleRTT - EstimatedRTT|(recommended β = 0.25 = 1/4) -
Retransmission Timeout Interval:
TimeoutInterval = EstimatedRTT + 4 × DevRTT
Exam Tip: Always compute in this exact order for EACH sample:
(1)EstimatedRTT→ (2)DevRTTusing the newly updatedEstimatedRTT→ (3)TimeoutInterval.
Solved Example 1
Given: Initial EstimatedRTT = 100 ms, DevRTT = 5 ms, α = 0.125, β = 0.25. Measured SampleRTT = 140 ms, 90 ms.
-
After SampleRTT = 140 ms:
EstimatedRTT = 0.875(100) + 0.125(140) = 87.5 + 17.5 = 105 msDevRTT = 0.75(5) + 0.25|140 - 105| = 3.75 + 8.75 = 12.5 msTimeoutInterval = 105 + 4(12.5) = 155 ms
-
After SampleRTT = 90 ms:
EstimatedRTT = 0.875(105) + 0.125(90) = 91.875 + 11.25 = 103.125 msDevRTT = 0.75(12.5) + 0.25|90 - 103.125| = 9.375 + 3.28125 = 12.65625 msTimeoutInterval = 103.125 + 4(12.65625) = 153.75 ms
Solved Practice Problem 2 (Very High Chance)
Given: Initial EstimatedRTT = 100 ms, DevRTT = 10 ms, α = 0.125, β = 0.25. Measured SampleRTT = 110 ms, 150 ms, 120 ms.
-
Sample 1 (110 ms):
EstimatedRTT = 0.875(100) + 0.125(110) = 101.25 msDevRTT = 0.75(10) + 0.25|110 - 101.25| = 7.5 + 2.1875 = 9.6875 msTimeoutInterval = 101.25 + 4(9.6875) = 140 ms
-
Sample 2 (150 ms):
EstimatedRTT = 0.875(101.25) + 0.125(150) = 107.34375 msDevRTT = 0.75(9.6875) + 0.25|150 - 107.34375| = 7.265625 + 10.6640625 = 17.9297 msTimeoutInterval = 107.34375 + 4(17.9296875) = 179.0625 ms
-
Sample 3 (120 ms):
EstimatedRTT = 0.875(107.34375) + 0.125(120) = 108.9258 msDevRTT = 0.75(17.9296875) + 0.25|120 - 108.92578125| = 16.2158 msTimeoutInterval = 108.92578 + 4(16.21582) = 173.79 ms
(Bonus — HTTP RTT Delay Formula if asked in Module 4: Non-persistent HTTP takes 2 × RTT + transmission time per object: 1 RTT for TCP handshake + 1 RTT for HTTP request/response. Persistent HTTP uses 1 RTT once for TCP + 1 RTT per object without pipelining, or 1 RTT for all referenced objects with pipelining.)
2. TCP Segment Structure (Diagram + Explanation)
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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Source Port # (16) | Destination Port # (16) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Sequence Number (32) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Acknowledgment Number (32) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Header| |U|A|P|R|S|F| |
| Length| Unused |R|C|S|S|Y|I| Receive Window (16) |
| (4) | (6) |G|K|H|T|N|N| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Internet Checksum (16) | Urgent Data Pointer (16) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Options (variable length) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Application Data (variable length) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Field Explanations:
- Source & Destination Port Numbers (16 bits each): Used for multiplexing and demultiplexing to identify sending and receiving application sockets.
- Sequence Number (32 bits): Byte-stream number of the first byte of data in this segment (counts bytes, not segments).
- Acknowledgment Number (32 bits): Sequence number of the next byte expected from the other side (cumulative ACK).
- Header Length (4 bits): Length of the TCP header in 32-bit words (minimum 5 words = 20 bytes when Options is empty).
- Unused (6 bits): Reserved for future use.
-
Flag Field (6 bits):
-
ACK: Indicates the acknowledgment number field is valid. -
RST,SYN,FIN: Connection setup (SYN), teardown (FIN), and reset/abort (RST). -
PSH: Requests receiver to push buffered data immediately to the application. -
URG: Indicates urgent data is present (end marked by Urgent Data Pointer).
-
- Receive Window (16 bits): Used for flow control — specifies the number of bytes the receiver is willing to accept (free buffer space).
- Internet Checksum (16 bits): Error detection covering header, data, and pseudo-header.
-
Urgent Data Pointer (16 bits): Points to the last byte of urgent data when
URG = 1. - Options (variable) & Application Data (variable): Optional parameters (e.g., MSS negotiation, window scaling) followed by payload data.
3. TCP Three-Way Handshake (Connection Establishment)
Client (Host 1) Server (Host 2)
| |
|--- Step 1: SYN=1, seq = x (CR) ---------------->|
| | (allocates buffers)
|<-- Step 2: SYN=1, ACK=1, seq = y, ack = x+1 ----|
| |
(allocates buffers) |
|--- Step 3: ACK=1, seq = x+1, ack = y+1 (+data)->|
| |
[ CONNECTION ESTABLISHED ]
Steps:
-
Step 1 (SYN): Client sends a TCP segment with
SYN = 1, no application data, and an initial sequence numberseq = x(CR(seq = x)). -
Step 2 (SYN-ACK): Server allocates TCP buffers/variables and replies with
SYN = 1, ACK = 1, its initial sequence numberseq = y, andack = x + 1. -
Step 3 (ACK): Client allocates buffers/variables and responds with
ACK = 1,ack = y + 1(may include payload data). Connection is now ESTABLISHED.
Why 3-Way Handshake (and not 2-Way)?
- Prevents old duplicate
CRsegments (delayed in the network from an earlier session) from opening a phantom connection:- When an old
CR(seq = x)arrives at Host 2, Host 2 repliesACK(seq = y, ACK = x). - Host 1 realizes it never initiated that
CRand sendsREJECT(ACK = y)(RSTin TCP), causing Host 2 to abort cleanly. - Even if an old duplicate
DATAalso replays, itsACK = zfails to matchy, so Host 2 rejects it.
- When an old
4. Transport-Layer Multiplexing & Demultiplexing
- Multiplexing (at sender): Gathering data chunks from different sockets at the source host, encapsulating each with a transport-layer header (source/dest ports), and passing segments to the network layer.
- Demultiplexing (at receiver): Examining header fields of incoming segments to identify the target socket and delivering data to the correct application process.
-
Port Numbers: 16-bit numbers (
0–65535). Ports0–1023are well-known ports (e.g., HTTP = 80, FTP = 21, SMTP = 25). The source port acts as a "return address" so the receiver knows where to send replies.
| Feature | Connectionless Demultiplexing (UDP) | Connection-Oriented Demultiplexing (TCP) |
|---|---|---|
| Socket Identifier |
2-tuple: (Destination IP, Destination Port)
|
4-tuple: (Source IP, Source Port, Destination IP, Destination Port)
|
| Behavior | Segments with the same dest IP & dest port from different sources go to the same socket. | Segments from different source IPs or source ports (even targeting port 80) go to separate sockets (one per connection). |
5. Differences Between IPv4 and IPv6
| Parameter | IPv4 | IPv6 |
|---|---|---|
| Address Size | 32-bit (~4.3 billion addresses, 2³²) | 128-bit (~3.4 × 10³⁸ addresses, 2¹²⁸) |
| Notation | Dotted decimal (192.168.1.10) |
Colon-separated hexadecimal (2001:0db8:85a3::8a2e:0370:7334) |
| Header Size | Variable 20–60 bytes (due to Options) | Fixed 40 bytes (8 fields; faster routing) |
| Header Checksum | Present (recomputed at every router) | Removed (handled by Link & Transport layers) |
| Fragmentation | Performed by both sender and routers | Performed only by sender (Path MTU Discovery) |
| Broadcast | Supported | Not supported (uses Multicast & Anycast) |
| Address Config | Manual or DHCP | Auto-configuration (SLAAC) + DHCPv6 |
| Security (IPsec) | Optional | Built-in support |
| NAT Dependence | Heavy reliance on NAT | Eliminates need for NAT |
(Transition Mechanisms: **1. Dual Stack, **2. Tunneling, **3. Protocol Translation / NAT64.)
6. Crash Recovery
- Problem: Packet loss on the wire is handled by retransmissions, but if a host crashes and restarts, it loses volatile connection state (sequence numbers, unacked buffers).
-
Why Crash Recovery Is Difficult (Stop-and-Wait Client → Server):
-
Case 1: ACK First, Then Write
- Server sends
ACKbefore delivering data to the application. - If server crashes after sending
ACKand before writing, the client assumes success → Result: Data is lost.
- Server sends
-
Case 2: Write First, Then ACK
- Server delivers data to the application before sending
ACK. - If server crashes after writing and before sending
ACK, the client retransmits → Result: Duplicate data is delivered.
- Server delivers data to the application before sending
-
Case 1: ACK First, Then Write
-
Takeaway: Without stable storage or application-level idempotency, a transport protocol cannot avoid both loss and duplication across a host crash. (Also note the **Forbidden Region* on restart: avoid reusing sequence numbers until old packets expire after time
T.)*
Module 4 — High-Priority Questions & Exam Answers
7. HTTP: Persistent vs Non-Persistent Connections (+ 2-Mark Quick Hits)
| Feature | Non-Persistent HTTP (HTTP/1.0) | Persistent HTTP (HTTP/1.1) |
|---|---|---|
| TCP Connections | Opens a new TCP connection for each object (HTML + each image). | Uses a single TCP connection left open to send multiple objects. |
| Connection Close | Server closes TCP connection right after sending one object. | Remains open until timeout or Connection: close. |
| Time / Overhead |
2 × RTT + transmit time per object (high latency & TCP overhead). |
1 RTT setup + 1 RTT per object (or 1 RTT total for all objects with pipelining). |
| Example (1 HTML + 10 JPEGs) | Needs 11 TCP connections (~22 RTT). | Needs 1 TCP connection (~12 RTT unpipelined, ~3 RTT pipelined). |
2-Mark HTTP Quick Hits:
- Why HTTP is stateless: Server stores no state/history about past client requests.
-
Conditional GET: Allows a Web cache (proxy) to verify cached content is fresh. Sends
GETwithIf-modified-since: <date>; if unchanged, server respondsHTTP/1.1 304 Not Modifiedwith an empty body. - Web Caching Benefits: (1) Reduces client response time; (2) Reduces traffic on the institution's Internet access link.
-
Status Codes:
200 OK,301 Moved Permanently,400 Bad Request,404 Not Found,505 HTTP Version Not Supported.
8. Cookies — How Information Is Stored & User-Server Interaction
Because HTTP is stateless, websites use cookies to identify returning users.
Four Components of Cookie Technology:
-
Set-cookie:header line in the HTTP response message (server → client). -
Cookie:header line in subsequent HTTP request messages (client → server). - A cookie file stored on the user's end system and managed by the browser (e.g.,
amazon: 1678,ebay: 8734). - A back-end database at the website mapping cookie ID (
1678) to user data/cart.
Step-by-Step Flow:
- Client sends an HTTP request to
amazon.com. - Server generates ID
1678, saves it in its back-end DB, and responds withSet-cookie: 1678. - Browser writes
amazon: 1678into the local cookie file. - On every later visit, the browser attaches
Cookie: 1678to the HTTP request header; the server queries its DB for personalized actions.
9. DNS — Function, Types of DNS Servers & Resource Records
-
Main Function: Directory service mapping human-friendly hostnames (e.g.,
www.bietdvg.edu) to IP addresses (121.7.106.83). Uses UDP port 53. - Why Distributed (not Centralized)? A centralized DNS suffers from: (1) Single point of failure, (2) Massive traffic volume, (3) High latency from distant DB, (4) Maintenance difficulty.
A. Types / Classes of DNS Servers
- Root DNS Servers: Top of hierarchy (13 root server identities A–M globally); return IP addresses of TLD servers.
-
Top-Level Domain (TLD) Servers: Handle generic TLDs (
.com,.org,.net,.edu,.gov) and country TLDs (.in,.uk,.jp); return the authoritative DNS server IP. - Authoritative DNS Servers: Maintained by organizations with public hosts (Web/mail servers); hold actual hostname-to-IP records.
- Local DNS Server: ISP/university resolver acting as proxy between client and the DNS hierarchy (8 messages total in a typical uncached lookup: 4 queries + 4 replies).
B. Types of DNS Resource Records (Name, Value, Type, TTL)
| Type | Name |
Value |
Purpose & Example |
|---|---|---|---|
A |
Hostname | IPv4 Address | Hostname-to-IP mapping: (relay1.bar.foo.com, 145.37.93.126, A)
|
NS |
Domain | Authoritative DNS hostname | Identifies authoritative name server: (foo.com, dns.foo.com, NS)
|
CNAME |
Alias hostname | Canonical hostname | Alias-to-real hostname: (foo.com, relay1.bar.foo.com, CNAME)
|
MX |
Domain | Mail server hostname | Identifies mail server for SMTP delivery |
(Query Types: Client → Local DNS is usually **recursive; Local DNS → Root → TLD → Authoritative is usually **iterative.)
10. Difference Between HTTP and SMTP
Both transfer files over TCP (HTTP port 80, SMTP port 25), with three key differences:
| Aspect | HTTP | SMTP |
|---|---|---|
| Mode / Direction | Pull Protocol: Users pull files from a Web server at their convenience (Server → Browser). | Push Protocol: Sending mail server pushes email messages to receiving mail server. |
| Data Format | Does not require 7-bit ASCII (supports binary directly). | Requires every message (including body) in 7-bit ASCII (uses MIME/Base64 for attachments). |
| Encapsulation | Encapsulates each object in its own HTTP response message. | Places all objects of a message into one single email message. |
(Mail Access Protocols to **pull* mail from recipient's mail server: POP3 [port 110; Authorization, Transaction, Update phases; +OK / -ERR] and IMAP [server-side folders, search, partial fetch].)*
11. SDN & AI in SDN (Module 4 — Part 2)
Software-Defined Networking (SDN) decouples the Control Plane (routing logic/software) from the Data Plane (packet-forwarding switches) and centralizes network intelligence in a programmable SDN Controller.
+-------------------------------------------------------+
| Application Plane (Routing, Firewall, AI Apps) |
+-------------------------------------------------------+
^ Northbound API (REST) v
+-------------------------------------------------------+
| Control Plane: SDN Controller + AI/ML Engine |
| (Global topology view, flow computation, policies) |
+-------------------------------------------------------+
^ Southbound API (OpenFlow) v
+-------------------------------------------------------+
| Data Plane: Switches & Routers (Forwarding Tables) |
+-------------------------------------------------------+
Three SDN Planes:
- Data Plane: Hardware switches/routers forwarding packets using flow tables.
- Control Plane: Centralized SDN Controller computing paths and pushing flow rules via Southbound API (OpenFlow).
- Application Plane: Traffic engineering, security, and monitoring apps talking to the controller via Northbound APIs.
Role of AI in SDN:
- Traffic Prediction & Load Balancing: ML predicts congestion and proactively reroutes flows.
- Intelligent Routing (Heuristic / RL): Dynamically optimizes latency, throughput, and jitter.
- Anomaly & Intrusion Detection: Detects DDoS/attacks in real time and pushes OpenFlow drop rules automatically.
- Intent-Based Networking: Translates high-level policies into flow rules.
12. Self-Healing Networks (Module 4 — Part 2)
A self-healing network automatically detects, diagnoses, isolates, and recovers from network faults, link cuts, or cyberattacks in real time with minimal or zero human intervention.
Closed-Loop Workflow:
- Monitor (Telemetry): Continuously gather link state, latency, packet loss, and router/switch metrics.
- Detect & Analyze (AI/ML): Spot hardware failures, congestion spikes, or security anomalies.
- Diagnose (Root Cause Analysis): Pinpoint failing nodes, links, or misconfigurations.
- Heal / Reconfigure: SDN controller computes backup paths, updates flow tables, isolates compromised nodes, or restarts failed services.
- Verify & Learn: Confirm SLA recovery and update models for faster future remediation.
Benefits: Maximum uptime/availability, near-zero MTTR (Mean Time to Recovery), reduced manual operations, and resilience against outages or DDoS attacks.
30/30 Exam Strategy
-
Nail the RTT Numerical first (
EstimatedRTT,DevRTT,TimeoutInterval) — guaranteed full marks. - Draw neat diagrams for TCP Segment Header (exact bit widths), 3-Way Handshake, Cookies, and SDN 3-Plane Architecture.
- Use comparison tables (IPv4 vs IPv6, Persistent vs Non-Persistent HTTP, HTTP vs SMTP, UDP vs TCP Demux) for crisp, examiner-friendly full-credit answers.
Top comments (0)