Computer Networks Previous Year Questions
Examining GATE computer networks questions from 2015 through 2024 reveals that IP subnetting and addressing appear in roughly 80 percent of papers, making ...
Patterns in GATE Computer Networks PYQs
Examining GATE computer networks questions from 2015 through 2024 reveals that IP subnetting and addressing appear in roughly 80 percent of papers, making it the single most reliable topic to study for guaranteed marks. Sliding window protocol efficiency calculations appear in about 70 percent of papers. TCP behavior (handshake, congestion control, flow control) and CRC error detection each appear in roughly 60 percent of papers.
Subnetting PYQs follow predictable structures. The most common variant provides a network address with prefix length and a requirement to divide it into a specified number of subnets or subnets of specified host capacity, asking for the subnet mask, number of usable hosts, or a specific subnet's broadcast address. A second common variant provides a list of IP addresses and asks which ones belong to the same subnet given a mask.
Sliding window PYQs ask for throughput or efficiency given bandwidth, frame size, propagation delay, and window size. They may also ask for the minimum sequence number space needed for a given window size under GBN versus SR. A variant asks for the maximum utilization achievable with a specified number of bits in the sequence number field.
TCP PYQs often trace the congestion window evolution through a sequence of events (connection establishment, data transfer, congestion signal, more data transfer) and ask for the cwnd value at a specific time or the total data transferred. You must apply slow start and congestion avoidance rules precisely at each event boundary.
Application layer PYQs tend to be conceptual: which protocol uses which port, what happens when a DNS query cannot be resolved, how HTTP persistent connections work, or what the purpose of SMTP AUTH is. These are often one-mark questions rewarding precise knowledge of protocol behavior.
Routing PYQs may ask you to run Dijkstra's algorithm on a small graph and find shortest path distances, or identify the issue with distance-vector routing (count-to-infinity) and how split horizon addresses it.
Solving Network Numericals from GATE PYQs
This section demonstrates worked solutions for high-frequency GATE CN numerical types, using the disciplined step-by-step approach that prevents errors under exam pressure.
Problem 1 — Subnetting. Given 172.16.0.0/12, divide into subnets requiring 1000 hosts each. Minimum host bits needed: ceiling of log2(1000 + 2) = 10 bits (2 raised to 10 = 1024 addresses, 1022 usable). Subnet prefix = 32 minus 10 = /22. Number of subnets = 2 raised to (22 minus 12) = 2 raised to 10 = 1024 subnets. First subnet: 172.16.0.0/22 (hosts 172.16.0.1 to 172.16.3.254, broadcast 172.16.3.255). Second subnet: 172.16.4.0/22. Each subnet has a block size of 4 in the third octet (2 raised to 2 = 4, since host bits extend into the third octet by 2 bits).
Problem 2 — Sliding window efficiency. Link: 1 Mbps bandwidth, 1000-bit frames, 270 ms one-way propagation delay, GBN with 7-frame window. Tt = 1000/1,000,000 = 1 ms. a = 270/1 = 270. 1 + 2a = 541. Window = 7 less than 541. Efficiency = 7/541 = 0.01294 = approximately 1.3 percent. Maximum throughput = 0.01294 times 1 Mbps = 12.94 kbps. To achieve 100 percent efficiency, window must be at least 541 frames.
Problem 3 — CRC. Message M = 1101011011, Generator G = 10011 (degree 4). Append 4 zeros: 11010110110000. Divide by 10011 using XOR (modulo-2 division): perform binary long division without carries. The 4-bit remainder is the FCS. Append FCS to message for transmission. At receiver, divide entire received frame by G; zero remainder means no error.
Problem 4 — TCP throughput. Sender window = 64 KB = 65536 bytes. RTT = 50 ms. Throughput = window/RTT = 65536/0.05 = 1,310,720 bytes/s = approximately 10.48 Mbps. If the actual link is 100 Mbps, the sender is limited by the window, not the link — this illustrates the importance of the bandwidth-delay product in determining required window size.
Tricky Network PYQ Concepts and How to Handle Them
Certain computer networks concepts generate disproportionate confusion in GATE PYQs because they involve subtle distinctions or counter-intuitive protocol behaviors. Mastering these tricky points is high-leverage for boosting your score.
Propagation delay versus transmission delay: these are independent quantities that students routinely conflate. Transmission delay (Tt) depends on frame size and bandwidth; propagation delay (Tp) depends on link length and propagation speed. A very short frame on a slow link has high Tt but possibly low Tp. A very large frame on a fast link has low Tt but potentially high Tp. In throughput calculations, both must be accounted for through the parameter a = Tp/Tt.
Store-and-forward delay at intermediate nodes: in a multi-hop network, each router must receive the entire packet before forwarding it. Total store-and-forward delay for n hops equals n times Tt (assuming zero queuing delay). However, with pipelining (when the next packet begins transmission before the first is fully forwarded), the total delay for k packets over n hops is n times Tt plus (k minus 1) times Tt.
Go-Back-N sequence number space: with n-bit sequence numbers, GBN allows window size at most 2 raised to n minus 1. If the window were 2 raised to n (the full space), the receiver cannot distinguish between a retransmission of old frames and new frames from the next window. Students often mistake GBN window size as 2 raised to n or confuse it with SR's 2 raised to (n-1).
Distance-vector count-to-infinity: when a link fails, nodes that believed they had a short path through the failed link incrementally increase their distance estimate, each claiming a path through the other. Split horizon prevents advertising a route back to the node from which it was learned. Poison reverse advertises the route back as infinite — more aggressively prevents count-to-infinity but does not eliminate it in all topologies.
HTTP persistent connections: in HTTP/1.0, each request opens a new TCP connection (three-way handshake overhead per request). HTTP/1.1 default is persistent — the TCP connection is reused for multiple requests. Pipelining (in HTTP/1.1) allows sending multiple requests before receiving responses, but head-of-line blocking limits benefit. HTTP/2 true multiplexing eliminates head-of-line blocking at the application layer.