D686 Operating Systems for Computer Scientists - Set 3 - Part 2

Correct Answer: Content hashing

Content hashing generates fingerprints of memory pages, identifying identical content across VMs for merging while tracking sharing through copy-on-write mechanisms effectively.
Option a is incorrect because pattern matching is too computationally expensive for real-time memory deduplication across potentially millions of pages in systems.
Option c is incorrect because address comparison wouldn't identify identical content at different addresses, missing deduplication opportunities across different VM spaces.
Option d is incorrect because size checking only identifies potentially identical pages but can't confirm content matches without examining actual page data.

Correct Answer: Unified management interface

Storage virtualization presents diverse storage devices through common abstraction layer, simplifying management of different vendors' hardware through single consistent interface for administrators.
Option a is incorrect because virtualization adds abstraction overhead potentially decreasing speed, though benefits like caching might offset this in practice sometimes.
Option c is incorrect because virtualization layers consume additional resources for abstraction, potentially increasing rather than reducing power consumption in systems typically.
Option d is incorrect because virtualization doesn't increase physical capacity; it organizes and presents existing storage more flexibly without adding space fundamentally.

Correct Answer: Remote kernel debugging

Remote kernel debugging connects debugger on separate machine to target kernel through serial or network, enabling breakpoints and inspection without disrupting debugger operation.
Option a is incorrect because user-mode debugging can't access kernel space where privileged code executes, unable to debug kernel components or drivers directly.
Option c is incorrect because print statements provide limited information and can't pause execution or inspect state interactively like proper debugging requires typically.
Option d is incorrect because core dumps capture post-crash state but don't enable interactive debugging with breakpoints and stepping through code execution dynamically.

Correct Answer: Virtual address spaces

Virtual address spaces provide hardware-enforced isolation where each tenant's processes operate in separate address spaces, preventing cross-tenant memory access completely through MMU.
Option a is incorrect because network segmentation isolates network traffic but doesn't prevent memory access between processes on the same physical machine directly.
Option c is incorrect because file permissions protect filesystem resources but don't isolate memory access between processes running on shared hardware concurrently present.
Option d is incorrect because user accounts provide identity separation but processes could still potentially access shared memory without address space isolation enforced.

Correct Answer: Priority-based preemption

Priority-based preemption immediately interrupts lower-priority tasks when critical safety tasks need execution, ensuring collision avoidance runs without delay for safety assurance.
Option a is incorrect because cooperative scheduling relies on tasks voluntarily yielding control, unacceptable for safety-critical systems requiring guaranteed immediate response capabilities fundamentally.
Option c is incorrect because time sharing gives equal CPU access to all tasks, not ensuring critical safety functions receive immediate priority during emergencies.
Option d is incorrect because voluntary yielding depends on task cooperation, inappropriate for safety systems where critical tasks must run immediately regardless.

Correct Answer: Kernel bypass

Kernel bypass technologies like DPDK allow userspace applications direct hardware access, eliminating kernel overhead for ultra-low latency packet processing in trading systems.
Option b is incorrect because system calls transition to kernel mode adding latency, exactly what high-frequency trading systems need to avoid for performance.
Option c is incorrect because device drivers operate in kernel space, adding the kernel processing overhead that increases latency for network packet handling unacceptably.
Option d is incorrect because virtual networking adds abstraction layers increasing latency, opposite of the direct hardware access needed for minimal latency requirements.

Correct Answer: Consistency guarantees

Byte-addressable persistent memory allows direct memory-style access to persistent data, requiring new consistency models since traditional block-based file system approaches don't apply.
Option a is incorrect because persistent memory improves access speed; the challenge is maintaining consistency with memory-style access patterns not block operations traditionally.
Option c is incorrect because capacity limits are hardware constraints not fundamental challenges to file system consistency models for persistent memory architectures specifically.
Option d is incorrect because wear leveling is implementation detail for longevity, not the fundamental consistency model challenge of byte-addressable persistent storage systems.

Correct Answer: State preservation

State preservation callbacks notify apps before suspension, allowing them to save critical data ensuring seamless restoration when users return to the application.
Option a is incorrect because background execution continues processing but doesn't specifically handle the save/restore cycle needed for app switching scenarios properly.
Option c is incorrect because memory locking prevents swapping but doesn't provide the state saving mechanism needed for handling app lifecycle transitions correctly.
Option d is incorrect because priority elevation affects scheduling order but doesn't enable state preservation during application switching and lifecycle management events.

Correct Answer: Thermal balancing

Thermal balancing considers server temperature and cooling capacity when scheduling tasks, preventing hotspots while optimizing cooling efficiency and preventing thermal throttling effectively.
Option a is incorrect because random distribution ignores thermal characteristics, potentially creating hotspots that increase cooling costs and risk thermal throttling significantly occurring.
Option c is incorrect because round-robin placement distributes tasks evenly by count not thermal load, missing opportunities for thermal optimization in placement decisions.
Option d is incorrect because first-fit allocation uses first available server without considering thermal state, potentially overheating servers with poor cooling capacity available.

Correct Answer: Connection scaling

Connection scaling challenges arise from managing thousands of simultaneous TCP connections for peer-to-peer blockchain networks, requiring optimized socket handling and memory management.
Option a is incorrect because while memory is involved, the specific challenge is scaling network connections not general memory management for blockchain applications fundamentally.
Option c is incorrect because disk throughput affects blockchain storage but the unique OS challenge is handling massive concurrent network connections for consensus protocols.
Option d is incorrect because display rendering is irrelevant for blockchain nodes which typically run headless, focusing on network and computation not graphics.

Correct Answer: Watchdog timer

Watchdog timers automatically reset systems when software fails to respond within specified intervals, ensuring recovery from kernel failures maintaining critical monitoring functions.
Option a is incorrect because backup batteries provide power redundancy but don't detect or recover from software failures affecting patient monitoring systems critically.
Option c is incorrect because error logging records problems for analysis but doesn't provide automatic recovery from kernel failures during critical operation periods.
Option d is incorrect because user alerts notify about problems but don't ensure continued operation when kernel fails, requiring automatic recovery mechanisms instead.

Correct Answer: Cost-aware eviction

Cost-aware eviction policies consider access frequency, object size, and retrieval cost when replacing cache entries, optimizing for bandwidth costs not just hit rates.
Option a is incorrect because random replacement ignores access patterns and costs, leading to suboptimal bandwidth usage and increased content delivery expenses unnecessarily incurred.
Option c is incorrect because oldest first (FIFO) doesn't consider access frequency or size, potentially evicting popular content while keeping rarely accessed items cached.
Option d is incorrect because newest first would remove recently cached items that might be frequently needed, increasing bandwidth costs from repeated retrievals required.

Correct Answer: Result verification

Quantum computation's probabilistic results require multiple runs and statistical verification, challenging OS assumptions about deterministic program execution and result handling traditionally expected.
Option a is incorrect because memory allocation for quantum programs follows similar patterns to classical programs; the challenge is handling probabilistic results not memory.
Option c is incorrect because power management for quantum computers involves cooling systems, but the OS challenge is managing probabilistic computation results fundamentally different.
Option d is incorrect because file storage remains deterministic regardless of computation type; the challenge is handling and verifying probabilistic quantum results properly.

Correct Answer: Write combining

Write combining aggregates multiple small writes into larger operations, reducing storage subsystem overhead and improving throughput for high-volume small file operations effectively.
Option b is incorrect because read caching speeds up reads but doesn't address the challenge of handling millions of small write operations efficiently occurring.
Option c is incorrect because defragmentation reorganizes existing data but doesn't help with the immediate challenge of processing numerous incoming small writes concurrently.
Option d is incorrect because compression reduces storage size but adds processing overhead, potentially worsening the problem of handling many small writes quickly.

Correct Answer: Frame synchronization

Frame synchronization ensures virtual content rendering aligns temporally with camera frames, preventing visual artifacts where digital objects appear disconnected from real-world anchors.
Option a is incorrect because network latency affects remote content delivery but local AR rendering primarily needs synchronization between camera and graphics pipeline directly.
Option c is incorrect because memory bandwidth affects data transfer rates but the critical challenge is temporal alignment not data throughput for AR applications.
Option d is incorrect because storage speed impacts asset loading but doesn't address the real-time synchronization needed between camera input and rendered overlays.

Correct Answer: Pre-warmed environments

Pre-warmed environments maintain initialized runtime containers ready for immediate function execution, eliminating startup overhead when requests arrive, achieving millisecond response times effectively.
Option a is incorrect because processor speed improvements alone can't account for order-of-magnitude differences between traditional and serverless container startup times achieved practically.
Option c is incorrect because kernel size affects boot time but serverless platforms use running environments not repeated kernel boots for each function invocation.
Option d is incorrect because networking improvements help request routing but don't address the container initialization time that dominates cold start latency fundamentally.

Correct Answer: Predictable scheduling

Predictable scheduling ensures tasks complete within guaranteed time bounds, essential for meeting strict millisecond deadlines in 5G edge computing scenarios requiring deterministic behavior.
Option a is incorrect because file caching improves average performance but doesn't provide the timing guarantees needed for strict millisecond deadline compliance required.
Option b is incorrect because interrupt coalescing reduces overhead by batching interrupts, potentially increasing latency which contradicts ultra-low latency requirements of 5G.
Option d is incorrect because memory compression saves space but adds processing time, potentially violating tight timing constraints in latency-critical edge applications fundamentally.

Correct Answer: All-reduce operations

All-reduce operations aggregate gradients across all GPUs then distribute results back, creating communication bottlenecks that limit distributed training scalability without optimization implemented.
Option b is incorrect because point-to-point messages between GPU pairs don't create the systematic bottleneck that all-reduce operations cause during gradient synchronization phases.
Option c is incorrect because broadcast sends data from one to all but training primarily needs bidirectional gradient aggregation not one-way data distribution patterns.
Option d is incorrect because scatter operations distribute different data pieces to recipients, not the gradient aggregation pattern that dominates training communication overhead observed.

Correct Answer: Performance overhead

Homomorphic encryption operations are computationally intensive, creating extreme performance overhead that operating systems must manage through resource allocation and scheduling optimization strategies.
Option a is incorrect because key management exists in all encryption systems; the unique challenge is the computational overhead of operating on encrypted data directly.
Option c is incorrect because network bandwidth requirements don't fundamentally change with homomorphic encryption; the challenge is local computation overhead not data transfer.
Option d is incorrect because user authentication remains separate from the computational challenges of processing homomorphically encrypted data within applications running locally.

Correct Answer: Geo-fencing policies

Geo-fencing policies restrict data movement based on geographic location, ensuring compliance with regulations requiring data to remain within specific jurisdictional boundaries mandated.
Option a is incorrect because encryption keys protect data confidentiality but don't enforce geographic restrictions on where data can be stored or processed physically.
Option c is incorrect because network firewalls control network access but can't ensure data residency when systems span multiple geographic regions inherently designed.
Option d is incorrect because access controls manage who can access data but don't restrict geographic location where data resides or gets processed.