RAKUB Maintenance Engineer (PO) Exam: 05.10.2021প্রশ্ন ও সমাধান

Multithreading vs Multitasking

Multitasking is the OS's ability to run multiple processes (independent programs) seemingly at the same time by rapidly switching the CPU between them (time-sharing) or actually running them in parallel on multiple cores.

Multithreading is the ability of a single process to run multiple threads (lightweight sub-units of execution) concurrently, all sharing the same address space, code, and data.

Aspect	Multitasking	Multithreading
Unit of execution	Process	Thread (within one process)
Memory	Each process has its own address space	Threads share the same address space
Context-switch cost	Higher (full process context)	Lower (only registers/stack of the thread)
Communication	Needs IPC (pipes, sockets, shared memory)	Direct - shared variables
Example	Running a browser and a music player at once	A browser using separate threads for rendering, networking, and UI

graph TD
  subgraph Multitasking
    P1[Process 1] -.CPU time-slice.-> CPU((CPU))
    P2[Process 2] -.CPU time-slice.-> CPU
  end
  subgraph Multithreading_in_one_process
    T1[Thread 1] --> SM[(Shared Memory)]
    T2[Thread 2] --> SM
  end

Exam takeaway: multitasking gives concurrency across programs; multithreading gives concurrency within one program - threads are cheaper to create and switch between than processes because they share memory.

Why RAID Is Necessary, and Choosing a Level for Large-Scale Data

RAID (Redundant Array of Independent Disks) combines multiple physical disks into one logical unit to achieve:

• Redundancy / fault tolerance - data survives a single disk failure (mirroring or parity).
• Performance - reads/writes can be striped and served in parallel across disks.
• Larger effective capacity and availability for a growing dataset.

Recommended level for an organization with a huge data workload: RAID 10 (1+0) - it stripes data across mirrored pairs of disks.

graph TD
  D1[Disk 1] ---|mirror| D2[Disk 2]
  D3[Disk 3] ---|mirror| D4[Disk 4]
  S[Data striped across both mirrored pairs] --> D1
  S --> D3

Why RAID 10 fits a huge-data organization

• Combines striping (RAID 0) for speed with mirroring (RAID 1) for safety, unlike RAID 5/6 which compute parity and suffer heavier write overhead.
• Fast rebuild: if a disk fails, data is simply copied from its surviving mirror - much faster than parity-based RAID 5/6 rebuilds, which must recompute the entire array (risking a second failure during a long rebuild window).
• Tolerates multiple disk failures as long as both disks of the same mirrored pair do not fail together.
• Trade-off: uses only 50% of raw capacity (versus roughly 75-90% for RAID 5/6), but a huge-data organization typically values speed and fast recovery over raw storage efficiency.

Exam takeaway: RAID 0 = speed only (no redundancy); RAID 1 = mirroring; RAID 5/6 = striping with parity (better capacity, slower writes/rebuilds); RAID 10 = best balance of speed and fast recovery for mission-critical, high-transaction workloads.