Lecture_13-CSCIU511-Jahangir_Majumder-Spring_2025

CSCI U511 01 Operating Systems

Instructor: AKM Jahangir A Majumder, PhD
Semester: Spring 2025
Lecture Date: February 27, 2025
Note: Some slides adapted from previous instructors; some figures from the textbook.

Finished coverage on Synchronization: Deadlock.
Continued discussion on Scheduling:
- Uniprocessor: FIFO, SJF, RR, MFQ
- Multiprocessor: Affinity and Gang
Quiz 3 grades and key posted on Blackboard.
Homework 4 due on Tuesday, March 4, 2025.
Quiz 4 scheduled for Thursday, March 6, covering lectures 11-13.
Sign-up link for Engage Green SPACE visit on Friday, March 21, is on Blackboard.
Mid-project progress report due Friday, February 28, at midnight on Blackboard.

Goal: Execute jobs in the order they arrive.
Fairness: Similar to a bank teller line.
Simplicity: Easy to implement.
Problem: Long jobs delay all subsequent jobs.
- Example Queue: A (4), B (3), C (6), D (3)
- Execution Order: A -> B -> C -> D.

Goal: Execute the task with the shortest remaining work.
Also known as Shortest Remaining Time First (SRTF).
Example: If five tasks arrive with varying lengths, SJF prioritizes shorter tasks.
- Example Completion:
  - FIFO: A (4), B (3), C (6), D (3)
  - SJF: B (3), D (3), A (4), C (6).
Challenge: How to accurately predict job length to prevent delays.

Policy: Non-preemptive selection of the shortest expected processing time.
Starvation Risk: Longer processes lag behind, potentially starving.
Accurate time estimation crucial; incorrect predictions may lead to job abortion.

Policy: Assign time slices (quantum) in a circular order.
- Preemption occurs if a task doesn't finish in its time slice.
Quantum Length: Balancing act—too short increases context switches; too long may delay response.
Typical quantum range: 10-50 ms.

Completion Times: Analyzed between FCFS, SJF, and RR in diverse scenarios.
Example:
- Process Arrival & Service Time:
  - A (3), B (6), C (4), D (5), E (2)
- Completion Time: Different algorithms yield varying results.

Balances CPU allocation among I/O-bound and compute-bound tasks.
Strategies:
- Maximize the minimum allocation.
- If any task requires less CPU, prioritize the least demanding.
- Split remaining resources equitably among tasks.

Assign a priority level to each process; highest priority runs first.
Dynamic priority adjustments based on CPU usage and I/O wait times.
Group processes into multiple queues by priority with round-robin scheduling per queue.

Definition: Extension of RR with multiple queues of varying priority levels.
Goals include:
- Responsiveness for short tasks, low overhead, starvation prevention, and fairness.
High-priority queues have short time slices; low-priority queues have longer intervals.

FIFO: Simple, low overhead but poor average response time with variable task sizes.
SJF: Optimal for average response time; poor for response time variance.
Round Robin: Poor response time with equal size tasks; benefits from max-min fairness strategies.

Challenges: Effective utilization of cores and parallel application adaptation.
Potential issues include contention for scheduler spinlocks, cache coherence overhead, and limited cache reuse.

Oblivious Scheduling: Each processor independently time-slices threads from its own ready list.
Ensures fair distribution of processing across applications.