Computer Systems Architecture FINAL CS

Serial Access

Data must be read sequentially; previous items need to be read or skipped to access the desired data.

Random Access

Data can be accessed directly without reading through other data, like in memory (RAM).

Primary Storage

The memory directly accessed by the CPU, such as RAM and cache memory.

Secondary Storage

Storage that is not directly accessed by the CPU; slower access times compared to primary storage.

Magnetic Disks

Disks with circular platters coated with magnetic material used for storing data.

Solid State Memory

Non-volatile storage using NAND flash chips, often used for secondary storage due to its speed and low power consumption.

Optical Disks

Storage media like CDs and DVDs that use lasers for reading and writing data.

Magnetic Tapes

Long-term storage using magnetic media, typically for backups, often stored in LTO (Linear Tape Open) format.

Parallel Advanced Technology Attachment (PATA)

Early hard disk interface using a parallel interface for data transfer, allowing up to two devices per channel.

Serial Advanced Technology Attachment (SATA)

Replaced PATA, using a serial interface for faster data transfer and smaller, more flexible cables.

Small Computer System Interface (SCSI)

Parallel interface that supports up to 30 devices, often used in high-capacity or performance applications.

Serial Attached SCSI (SAS)

Combines the advantages of SCSI and serial technology, offering better performance, scalability, and reduced conflicts.

Access Time

The total time taken to start transferring data, including seek time and rotational delay.

Seek Time

The time taken for the read/write head to move to the correct track on the disk.

Rotational Delay

The time it takes for the required disk sector to rotate under the read/write head.

Rotation Speed

The speed at which the disk platter spins, measured in revolutions per minute (RPM).

File System

The method of organizing and storing data on storage devices, such as FAT, NTFS, and EXT.

RAID (Redundant Array of Independent Disks)

A method of using multiple disks to increase performance, redundancy, and fault tolerance.

Mirrored Array

A RAID configuration where data is duplicated across multiple disks for redundancy.

Striped Array

A RAID configuration that divides data into blocks and writes it to multiple disks simultaneously for improved performance.

Virtualization

A technology that allows one physical computer to act as multiple virtual machines, each with its own operating system and resources.

Preemptive Multitasking

The OS allocates specific time slices to tasks and can interrupt tasks to switch to another one.

Concurrency

The perception of tasks running at the same time, even if they're not literally executing simultaneously.

Parallelism

Tasks literally running at the same time, using multiple processors or cores.

Data Race

Occurs when one thread modifies shared data while another is reading it, leading to corrupted results.

Race Condition

When multiple threads try to access and modify the same data simultaneously, causing unpredictable behavior.

Synchronization

Ensures that only one thread can access shared resources at a time, preventing data corruption.

Atomic Operations

Operations that are completed entirely without interruption, ensuring data integrity in multithreading scenarios.

Immutable Data

Data that cannot be modified, ensuring safe reading across multiple threads without conflicts.

Multicore Microprocessor

A processor that contains multiple processing units (cores) within a single integrated circuit.

Task-Level Parallelism

Running independent programs simultaneously across multiple processors.

Flynn's Taxonomy

A classification of parallel computing systems based on instruction and data streams: SISD, SIMD, MISD, MIMD.

SISD (Single Instruction, Single Data)

A non-parallel system where one instruction is executed on one data stream at a time.

SIMD (Single Instruction, Multiple Data)

A parallel system where one instruction is executed across multiple data elements, suitable for graphics and image processing.

MISD (Multiple Instruction, Single Data)

A parallel system where multiple instructions operate on the same data stream, rarely used in practice.

MIMD (Multiple Instruction, Multiple Data)

A parallel system where multiple processors execute different instructions on different data streams, common in modern supercomputers.

Multithreading

Running multiple threads within a single processor, allowing better utilization of hardware resources.

Fine-Grained Multithreading

Switching between threads at each instruction to interleave execution and maximize processor utilization.

Coarse-Grained Multithreading

Switching threads only on costly stalls, such as cache misses, to optimize processing.

Simultaneous Multithreading (SMT)

A technique that allows multiple threads to run simultaneously on a single processor, improving throughput (e.g., Intel's Hyperthreading).

Vector/Array Processing

Using vector registers and pipelined execution units to process data elements in parallel.

High-Performance Computing (HPC)

The use of aggregated computing resources to solve data-intensive tasks, such as simulations and modeling.

Beowulf Clusters

Low-cost, highly configurable clusters of computers connected via private Ethernet for parallel processing tasks.

Grid Computing

A method of distributing computing tasks across a network of computers to achieve supercomputing performance.

Cloud Computing

The use of remote servers and services over the Internet to store, manage, and process data instead of using local devices.

Software as a Service (SaaS)

A cloud-based software delivery model where applications are hosted by a provider and accessed over the internet.

Infrastructure as a Service (IaaS)

Cloud service providing virtualized computing resources, including virtual machines, storage, and networking.

GPU Architecture

A processor optimized for handling 2D/3D graphics, video, and visual computing tasks.

CUDA

A parallel programming platform and model developed by NVIDIA for leveraging GPU power for general-purpose computing.

Direct3D

A Microsoft API for rendering 3D graphics in applications, primarily for games and multimedia.

OpenGL

An open standard API for rendering 2D and 3D vector graphics used in gaming and simulations.

OpenCL

A framework for writing programs that execute across heterogeneous platforms, including GPUs and CPUs.

Vulkan

A cross-platform API for graphics and computing, providing lower-level control over GPU resources.