ISRO Scientist/Engineer Computer Science 2025 paper: Part A, questions 21 to 40, solved
Questions 21 to 40 of the ISRO Scientist/Engineer 'SC' Computer Science written test of 26 Oct 2025 (Set A): 20 questions, each solved by Skill-Lync and checked against ISRO's final answer key.
Question paper and answer keys published by ISRO's Centralised Recruitment Board; solutions by Skill-Lync · Updated 18 Nov 2026 · 20 of 80 questions on this page
What elasticity in cloud computing refers to
- (a)Ability to scale up and scale down of compute resources
- (b)Parallel execution of jobs
- (c)High availability of systems
- (d)Minimum response time
Answer: (a) Ability to scale up and scale down of compute resources Official final key: (a)
Elasticity means resources are added when demand rises and released when it falls, automatically, so you pay for what the load needs.
Parallel execution is concurrency, high availability is about staying up despite failures, and minimum response time is performance. None of these is elasticity.
(a) Ability to scale up and scale down of compute resources
Auto-scaling groups let a data-processing pipeline take on many servers while a satellite pass is being downloaded and release them afterwards.
Matching bare metal and hosted hypervisors to their descriptions
(P) Runs on host machine physical hardware
(Q) Hosted hypervisor
(R) Runs on the operating system of physical host machine
(S) Bare metal hypervisor
Which of the following is correct match?
- (a)Type-1 Hypervisor : (P) and (Q), Type-2 Hypervisor : (R) and (S)
- (b)Type-1 Hypervisor : (P) and (S), Type-2 Hypervisor : (Q) and (R)
- (c)Type-1 Hypervisor : (R) and (S), Type-2 Hypervisor : (P) and (Q)
- (d)Type-1 Hypervisor : (R) and (Q), Type-2 Hypervisor : (P) and (S)
Answer: (b) Type-1 Hypervisor : (P) and (S), Type-2 Hypervisor : (Q) and (R) Official final key: (b)
A Type-1 hypervisor runs directly on the physical hardware, which is why it is called bare metal (examples: Xen, ESXi, KVM-based hosts). So Type-1 gets (P) and (S).
A Type-2 hypervisor is an ordinary program running on a host operating system, which is why it is called hosted (examples: VirtualBox, VMware Workstation). So Type-2 gets (Q) and (R).
This is option (b).
(b) Type-1: (P) and (S); Type-2: (Q) and (R)
Mission-control and cloud data centres run Type-1 hypervisors for server consolidation; developers use Type-2 ones for test machines on a laptop.
Which is not a characteristic of Big Data
- (a)Velocity
- (b)Visualisation
- (c)Volume
- (d)Variety
Answer: (b) Visualisation Official final key: (b)
Big Data is classically described by the V's: Volume (how much), Velocity (how fast it arrives) and Variety (many formats); Veracity and Value are often added. Visualisation is something we do with data after processing, not a property of the data.
(b) Visualisation
Telemetry and earth-observation archives are big in all three senses: volume of images, velocity of downlinked streams and variety of formats.
Primary purpose of HDFS in Big Data storage
- (a)To compress files
- (b)To store in-memory data
- (c)To store relational data
- (d)To store large files across multiple machines
Answer: (d) To store large files across multiple machines Official final key: (d)
HDFS splits each large file into blocks (128 MB by default in current Hadoop), stores the blocks on many commodity machines and replicates them for fault tolerance.
It is not a compression tool, not an in-memory store and not a relational database.
(d) To store large files across multiple machines
Archives of raw satellite passes or logs are stored on distributed file systems like HDFS so batch jobs can read them in parallel.
Big Data tool for real-time stream processing
- (a)Hadoop
- (b)Apache Flink
- (c)Splunk
- (d)ERDPlus
Answer: (b) Apache Flink Official final key: (b)
Apache Flink is a stream-processing engine: it handles unbounded data record by record with low latency. Hadoop (MapReduce) is built for batch jobs, Splunk is a log search and analytics product, and ERDPlus is a tool for drawing entity-relationship diagrams.
(b) Apache Flink
Stream engines process live telemetry or event feeds for alerting without waiting for a nightly batch.
Storing unstructured data in a Big Data environment
- (a)In-memory database
- (b)Relational database
- (c)NOSQL database
- (d)SQL database
Answer: (c) NOSQL database Official final key: (c)
Relational (SQL) databases need a fixed schema of tables and columns. Unstructured and semi-structured data (documents, logs, JSON, text) has no such fixed shape, so NoSQL stores (document, key-value, column-family, graph) are used. An in-memory database describes where data is held, not its shape.
(c) NOSQL database
Document stores hold varied event and log records whose fields change from one mission or instrument to the next.
Matching structured, semi-structured and unstructured data
| (i) | Structured data | (P) | Text files, Images, Videos |
| (ii) | Semi-structured data | (Q) | Fixed format data |
| (iii) | Unstructured data | (R) | XML, JSON |
- (a)(i)-(Q), (ii)-(R), (iii)-(P)
- (b)(i)-(Q), (ii)-(P), (iii)-(R)
- (c)(i)-(P), (ii)-(R), (iii)-(Q)
- (d)(i)-(R), (ii)-(Q), (iii)-(P)
Answer: (a) (i)-(Q), (ii)-(R), (iii)-(P) Official final key: (a)
Structured data has a fixed format (rows and columns of a table), so (i)-(Q). Semi-structured data carries its own tags or keys but no rigid table, as in XML and JSON, so (ii)-(R). Unstructured data has no predefined model: text files, images and videos, so (iii)-(P).
(a) (i)-(Q), (ii)-(R), (iii)-(P)
Pipelines routinely mix all three: a database table of observations, JSON metadata, and raw image files.
Hexadecimal constant matching the regular expression 0x[0-9A-F]+
(0x[0-9A-F]+)- (a)0x4C5
- (b)0x50b
- (c)04A
- (d)0x
Answer: (a) 0x4C5 Official final key: (a)
The expression needs the literal 0x followed by one or more characters from 0-9 or capital A-F.
0x4C5:0xthen 4, C, 5, all allowed. Matches.0x50b: lowercasebis not inA-F. No match.04A: nox. No match.0x: nothing after the prefix, but+needs at least one digit. No match.
(a) 0x4C5
Parsers for register dumps, memory maps and log formats rely on exactly this kind of pattern, where case and the minimum digit count matter.
Which statement is true for a virtual machine like JVM
- (a)Hardware dependent
- (b)Hardware independent
- (c)Bytecode is part of VM
- (d)Links object code
Answer: (a) Hardware dependent Official final key: (a)
A JVM is itself a native program (interpreter plus JIT compiler) written for one operating system and processor, so a separate JVM is needed for each platform: the VM is hardware dependent. What is portable is the Java bytecode it executes, which is why Java programs run unchanged on any machine that has a JVM.
Option (b) is the common loose statement about Java, but it describes the bytecode, not the VM. Bytecode is the VM's input, not a part of it, and linking object code is the job of a linker, not of the JVM.
(a) Hardware dependent
Embedded teams pick a VM or runtime build per target board, even though the application bytecode stays the same.
Compiler phase that receives tokenized output
- (a)Lexical Analyser
- (b)Parser
- (c)Pre-processor
- (d)Assembler
Answer: (b) Parser Official final key: (b)
The lexical analyser reads characters and produces tokens. For x = y + 12; it emits x, =, y, +, 12, ;. The next phase, the parser (syntax analyser), receives this token stream and checks it against the grammar.
The pre-processor runs before, on raw text, and the assembler works much later, on assembly code.
(b) Parser
Config-file and command-language parsers in ground software follow the same tokenise-then-parse split.
Tool that generates a parser from BNF notation
- (a)lex
- (b)yacc
- (c)ed
- (d)gdb
Answer: (b) yacc Official final key: (b)
yacc (Yet Another Compiler Compiler) takes grammar rules written in a BNF-like notation and generates a parser, usually an LALR(1) one. lex generates the lexical analyser (token scanner), ed is a line text editor, and gdb is a debugger.
(b) yacc
yacc and its descendant bison generate the parsers for many language tools and command interpreters.
Properties of an ambiguous grammar for the same input string
(i) two different parse trees
(ii) two leftmost derivations
(iii) two rightmost derivations
(iv) leftmost and rightmost derivations are different
Which one is correct?
- (a)(i), (ii), (iii) are true; (iv) is false
- (b)(i) is true; (ii), (iii), (iv) are false
- (c)(i), (iv) are true; (ii), (iii) are false
- (d)(i) is false; (ii),(iii),(iv) are true
Answer: (a) (i), (ii), (iii) are true; (iv) is false Official final key: (a)
A grammar is ambiguous if some string has more than one parse tree. A parse tree corresponds to exactly one leftmost and one rightmost derivation, so two trees mean two leftmost derivations and two rightmost derivations. Statements (i), (ii) and (iii) are true. For example, E -> E+E | a gives two trees for a+a+a; each of those 2 trees has its own leftmost and its own rightmost derivation.
Statement (iv) does not signal ambiguity. Even in an unambiguous grammar such as S -> AB, A -> a, B -> b, the leftmost derivation of ab (S, AB, aB, ab) differs from the rightmost one (S, AB, Ab, ab), so a difference between the two derivations is not a mark of an ambiguous grammar. So (iv) is not a valid consequence of ambiguity, and it is false.
(a) (i), (ii), (iii) are true; (iv) is false
Compiler writers remove ambiguity (operator precedence, dangling else) so every program has one meaning; parser generators report such conflicts.
Which string is generated by S to A | B, A to aAb | c, B to bBd | e
- (a)accbb
- (b)aacbb
- (c)abaed
- (d)aaebb
Answer: (b) aacbb Official final key: (b)
Work out the language. goes to or to , so the strings come from one of two separate families.
wraps the same number of 's on the left and 's on the right around a single . That gives .
does the same with and around an . That gives .
So with .
Check the options. (a) accbb has two 's, so no. (b) aacbb is , so yes: . (c) abaed mixes the two families, so no. (d) aaebb has in the middle but and around it, so no.
(b) aacbb
Context-free grammars like this define the syntax that parser generators (yacc, ANTLR) turn into code for command languages, telemetry-frame formats and configuration files.
Which statement can crash after ptr = new int[4] in C++
int *ptr;
ptr = new int [4];Which of the following statements may make the program crash when executed subsequently?- (a)
ptr++; - (b)
ptr += 3; - (c)
ptr[4] = 0x16; - (d)
ptr = null;
Answer: (c) ptr[4] = 0x16; Official final key: (c)
new int [4] gives an array whose valid indices are 0, 1, 2 and 3. ptr points at element 0.
(a) ptr++ moves the pointer to element 1, still inside the array. (b) ptr += 3 moves it to element 3, the last valid one. Neither touches memory outside the block.
(c) ptr[4] = 0x16 means *(ptr + 4) = 0x16. Index 4 is one element past the end, so this writes outside the allocation. That is undefined behaviour: it can silently corrupt the heap's bookkeeping or neighbouring data, and it may crash the program, which is exactly what the question asks ("may make the program crash").
(d) null is not a C++ keyword (the null pointer is nullptr, or the old macro NULL), so it would fail to compile rather than crash. Even ptr = nullptr only drops the pointer, leaking the four ints, and does not itself crash.
(c) ptr[4] = 0x16;
Out-of-bounds writes on heap arrays are one of the most common causes of crashes and security bugs in C and C++ code, including embedded flight and ground software, which is why bounds-checked containers and sanitisers are used.
OSI layer associated with a wrong default gateway setting
- (a)Layer 2
- (b)Layer 5
- (c)Layer 3
- (d)Layer 4
Answer: (c) Layer 3 Official final key: (c)
The default gateway is the IP address of the router a host sends packets to when the destination is outside its own subnet. Choosing the next hop from an IP address is routing, and routing belongs to the network layer, Layer 3.
Layer 2 (data link) deals with MAC addresses on the local link, Layer 4 (transport) with TCP and UDP ports, and Layer 5 (session) with dialogue control. A wrong gateway leaves the local link and ports fine, but packets for outside networks never reach the right router.
(c) Layer 3
Misconfigured gateways and routes are a routine first check in network troubleshooting, in data centres and in ground-station LANs alike.
IPSec configuration for confidentiality, integrity and hidden private IPs
- (a)IPSec using AH in Tunnel mode
- (b)IPSec using ESP in Transport mode
- (c)IPSec using ESP in Tunnel mode
- (d)IPSec using AH in Transport mode
Answer: (c) IPSec using ESP in Tunnel mode Official final key: (c)
Three needs: confidentiality of the data, integrity, and hiding the private IP addresses.
AH (Authentication Header) gives integrity and authentication only. It does not encrypt anything, so (a) and (d) fail on confidentiality.
ESP (Encapsulating Security Payload) encrypts and can also authenticate. In transport mode it protects only the payload and keeps the original IP header, so the private addresses stay visible. That rules out (b).
In tunnel mode ESP encrypts the whole original IP packet, header included, and wraps it in a new outer IP header (between the two office gateways). The private addresses are hidden inside the encrypted part.
(c) IPSec using ESP in Tunnel mode
Site-to-site VPNs between offices or between a ground station and a data centre are usually built exactly this way, ESP in tunnel mode between two gateways.
Nearest-codeword decoding of 01000 among 00000, 01110, 11000
Codeword A: 00000
Codeword B: 01110
Codeword C: 11000
During operation, receiver gets the word 01000. What would the system conclude?
- (a)It will correct to codeword A
- (b)It will correct to codeword B
- (c)It will correct to codeword C
- (d)It will be unable to correct the error
Answer: (d) It will be unable to correct the error Official final key: (d)
Nearest-codeword decoding picks the valid word with the smallest Hamming distance (number of differing bits) from the received word 01000.
To A (00000): one bit differs, distance 1. To B (01110): bits 3 and 4 differ, distance 2. To C (11000): only the first bit differs, distance 1.
A and C tie at distance 1, so there is no unique nearest codeword. The code itself explains why: the minimum distance between codewords is 2 (A to C is 2), and a code with minimum distance 2 can detect one error but cannot correct it.
The receiver knows the word is wrong but cannot tell whether A or C was sent.
(d) It will be unable to correct the error
Error-detecting and error-correcting codes protect telemetry and storage; a code's minimum distance decides how many bit errors it can detect or correct.
RSA public exponent e for p = 11, q = 13 and private key d = 7
- (a)103
- (b)143
- (c)21
- (d)19
Answer: (a) 103 Official final key: (a)
and .
The keys must satisfy , with coprime to .
Test the options with : , so works. For : , no. For : , no. For : , so it cannot be a public exponent.
, option (a)
Choosing and validating RSA key components (coprimality with , modular inverse) is what crypto libraries do at key generation; real keys use primes of 1024 bits or more.
Effect of blocking all incoming ICMP packets on the TCP/IP stack
- (a)No effect
- (b)Path MTU discovery fails
- (c)Slow start fails
- (d)Congestion control fails
Answer: (b) Path MTU discovery fails Official final key: (b)
Path MTU discovery sends packets with the Don't Fragment bit set. When one is too large for a link, the router drops it and returns an ICMP "destination unreachable, fragmentation needed" message (type 3, code 4) carrying the link's MTU. If the host's firewall blocks that incoming ICMP, it never learns the smaller size and keeps sending packets that are dropped. This is the well-known "PMTU black hole".
Slow start and congestion control are driven by TCP's own acknowledgements and timeouts, not by ICMP. So (c) and (d) are not affected, and (a) is wrong because (b) is.
(b) Path MTU discovery fails
Over-aggressive ICMP filtering is a classic cause of connections that hang on large transfers while small ones work, often seen with VPNs and tunnels.
Ideal TCP window for 8192000 bps and 32 ms latency
- (a)65536 bytes
- (b)8192 bytes
- (c)16384 bytes
- (d)32768 bytes
Answer: (d) 32768 bytes Official final key: (d)
To keep the link busy the sender must have a full bandwidth-delay product of data in flight, so the window equals bandwidth times delay.
Divide by 8 to get bytes:
Here "latency" is read as the round-trip time, so the window needed to keep the pipe full is the bandwidth-delay product over 32 ms, 32768 bytes. If 32 ms were the one-way delay, the round trip would be 64 ms and the window 65536 bytes (option (a)); ISRO’s key takes the single figure as the round trip.
(d) 32768 bytes
Sizing TCP windows to the bandwidth-delay product is central for long-latency links such as satellite or deep-space ground links, where default windows leave the pipe mostly empty.