Operating Systems

Operating Systems Today: From the Laptop to the Edge–Cloud

what an OS does (user and system views), the OS as resource manager, evolution from batch to cloud and edge, types of OS, multiprogramming and CPU utilization, goals of an OS, generic components, Amdahl's law, the device–edge–cloud continuum, the EdgeCampus case

Hardware Support, System Calls and OS Architecture

interrupts, traps and exceptions, interrupt handling and priorities, programmed I/O vs. interrupt-driven I/O vs. DMA, device controllers and drivers, dual-mode operation and privileged instructions, memory and CPU protection (base/limit, timer), system calls in Linux (fork, execve, open, read, write, kill), the cost of system calls and buffering, booting, monolithic, layered, microkernel, hybrid and unikernel architectures

Processes, Threads and Resource Management

the OS as resource manager, resource types (preemptable, sharable, reusable, consumable), goals of resource management, program vs. process, process memory layout, process control block, process states and transitions (including suspended states), context switching and its cost, fork/exec/wait/exit, zombies and orphans, threads, user vs. kernel threads, threading models, thread pools, Java threads

CPU Scheduling I: Classic Algorithms and Metrics

CPU–I/O burst cycle, scheduling levels (long-, medium-, short-term), preemptive vs. non-preemptive, criteria (utilization, throughput, turnaround, waiting, response), FCFS and the convoy effect, SJF and SRTF, burst prediction by exponential averaging, priority scheduling, starvation and aging, Round Robin and the time quantum, HRRN, multilevel and multilevel feedback queues, lottery and fair-share scheduling, Gantt charts

CPU Scheduling II: Multicore, Real-Time and Linux

asymmetric vs. symmetric multiprocessing, global vs. per-core run queues, load balancing (push/pull migration), processor affinity, NUMA, SMT, heterogeneous cores (big.LITTLE, P/E cores), gang scheduling, Linux scheduling classes, CFS and vruntime, EEVDF, nice weights, real-time tasks, utilization, rate-monotonic scheduling and the Liu–Layland bound, response-time analysis, EDF, priority inversion (Mars Pathfinder), priority inheritance, energy-aware scheduling and DVFS

Scheduling on Heterogeneous Systems: Workflows, HEFT and Beyond

scientific and AI workflows as DAGs, heterogeneous processors (CPU, GPU, edge, cloud), computation and communication costs, makespan, SLR, speedup and efficiency, NP-hardness, list scheduling, upward and downward ranks, HEFT with insertion, CPOP and the critical path, PEFT and the optimistic cost table, look-ahead scheduling (PPTS, IPPTS), budget- and deadline-aware scheduling, learning-based scheduling, how scheduling research is evaluated

Concurrency and Synchronization

race conditions and lost updates, interleavings, the critical-section problem (mutual exclusion, progress, bounded waiting), Peterson's solution, atomic hardware instructions (test-and-set, compare-and-swap), spinlocks and mutexes, semaphores, producer–consumer, readers–writers, dining philosophers, monitors and condition variables, Java synchronization (synchronized, locks, atomics, BlockingQueue, volatile), message passing, synchronization at cloud scale

Deadlocks: Prevention, Avoidance, Detection and Recovery

the four necessary conditions, resource-allocation graphs and wait-for graphs, the ostrich approach, deadlock prevention (breaking each condition, resource ordering), deadlock avoidance (safe states, the Banker's algorithm), deadlock detection with multiple instances, recovery (process termination, preemption, victim selection, rollback), livelock and starvation, deadlocks in Java, databases and distributed systems, gang scheduling of distributed GPU jobs

Memory Management: From Address Spaces to Paging

logical vs. physical addresses, address binding, the MMU with base and limit registers, dynamic linking and shared libraries, contiguous allocation (first, best, worst and next fit), external and internal fragmentation, compaction, paging (page number and offset), page tables and page-table entries, the TLB and effective access time, multi-level, hashed and inverted page tables, huge pages, segmentation, kernel allocators (buddy system, slab), memory limits in containers and the OOM killer, memory budgets of AI models at the edge

Virtual Memory: Demand Paging, Replacement and Thrashing

demand paging, page faults and their handling, effective access time with page faults, copy-on-write and fork, page replacement (FIFO, optimal, LRU, clock/second chance, enhanced clock), Belady's anomaly and stack algorithms, frame allocation (equal, proportional, global vs. local), thrashing, the working-set model and page-fault frequency, memory-mapped files, swap and compressed memory, Linux reclaim (kswapd, active/inactive lists), live migration of VMs and containers, paged KV caches for AI inference

File Systems, Disks and Storage

files, directories and links, the open-file table, file-system layout, allocation methods (contiguous, linked/FAT, indexed/inodes), free-space management (bitmaps), the page cache and write-back vs. write-through, fsync, crash consistency (journaling, copy-on-write file systems), hard-disk performance (seek, rotation, transfer), disk scheduling (FCFS, SSTF, SCAN, C-SCAN, LOOK, C-LOOK), SSDs (flash translation layer, write amplification, TRIM, wear leveling), NVMe, RAID levels, distributed and cloud storage (object storage, replication, erasure coding), data placement at the edge

Virtualization, Containers and Cloud Resource Management

virtual machines and hypervisors (type 1 and type 2), trap-and-emulate, hardware virtualization support, paravirtualization, memory virtualization (nested/extended page tables), I/O virtualization, overcommitment, containers (namespaces, cgroups, images and layers), VMs vs. containers, Kubernetes scheduling (filtering and scoring, requests and limits, QoS classes), CPU limits and CFS throttling, VM placement as bin packing (FF, BF, FFD, BFD), consolidation and energy, autoscaling (the Kubernetes HPA rule), serverless computing and cold starts, noisy neighbours and multi-tenancy

Edge Computing and Computation Offloading

why edge computing (latency, bandwidth, privacy, autonomy), the device–edge–cloud continuum, MEC, cloudlets and fog, the offloading decision (time and energy models, Shannon capacity, break-even bandwidth), DVFS and energy, partial offloading and DNN partitioning, multi-user offloading and congestion (game-theoretic view, price of anarchy), queueing at edge servers, deadline- and cost-aware placement across device, edge and cloud, workflow (DAG) offloading, mobility and service migration, edge orchestration (K3s, KubeEdge), research methods (optimization, Lyapunov, heuristics, reinforcement learning)

Protection and Security: From File Permissions to Zero-Trust Edge

security goals (confidentiality, integrity, availability), protection vs. security, the principle of least privilege, protection domains and the access matrix, access control lists vs. capabilities, Unix permissions, umask, setuid and Linux capabilities, mandatory access control (Bell–LaPadula, Biba, SELinux, AppArmor), authentication (password hashing, salts, slow hashes, multi-factor), memory-safety attacks and OS defenses (stack canaries, NX, ASLR), sandboxing (seccomp, containers), secure boot and TPM, encryption at rest and in transit, trusted execution environments, ransomware and supply-chain threats, securing edge and cloud systems (zero trust, updates, physical access)