💾Intro to Computer Architecture Unit 6 – Input/Output Systems in Computer Architecture

Key Concepts

• Input/Output (I/O) refers to the communication between a computer system and external devices (keyboards, mice, displays, storage devices, network interfaces)
• I/O performance significantly impacts overall system performance as I/O operations are often slower than CPU and memory operations
• I/O devices are connected to the computer system through various interfaces (USB, PCIe, SATA, Ethernet)
  • These interfaces define the physical and logical connection between the device and the system
• I/O operations are managed by the operating system through device drivers, which provide a standard interface for applications to interact with I/O devices
• I/O performance can be improved through techniques such as buffering, caching, and direct memory access (DMA)
• Interrupt-driven I/O allows the CPU to perform other tasks while waiting for I/O operations to complete, improving system efficiency
• Memory-mapped I/O and port-mapped I/O are two common methods for the CPU to communicate with I/O devices

I/O Basics and Terminology

• Input devices (keyboards, mice, scanners) send data to the computer system for processing
• Output devices (displays, printers, speakers) receive data from the computer system and present it to the user
• Storage devices (hard drives, SSDs, USB drives) provide persistent storage for data and programs
• Network interfaces (Ethernet, Wi-Fi) enable communication between computer systems over a network
• Ports are connection points for I/O devices, identified by unique addresses or names
• Bandwidth refers to the maximum amount of data that can be transferred per unit of time, typically measured in bits per second (bps) or bytes per second (B/s)
• Latency is the time delay between the initiation of an I/O operation and its completion, often measured in milliseconds (ms) or nanoseconds (ns)
  • Lower latency indicates faster response times and better performance

I/O Hardware Components

• I/O controllers are specialized hardware components that manage the communication between the CPU and I/O devices
  • Examples include USB controllers, SATA controllers, and network interface cards (NICs)
• I/O adapters are expansion cards that provide additional I/O functionality to a computer system (sound cards, graphics cards, network adapters)
• Buses are communication pathways that connect I/O devices to the CPU and memory (PCIe, USB, SATA)
  • Buses have specific protocols and architectures that define how data is transferred
• Connectors and cables physically connect I/O devices to the computer system (USB connectors, Ethernet cables, HDMI cables)
• I/O interfaces define the electrical, mechanical, and logical characteristics of the connection between I/O devices and the computer system (USB, PCIe, SATA, Ethernet)
• Sensors and actuators are I/O devices that interact with the physical world (temperature sensors, motion sensors, motors, relays)
• Human interface devices (HIDs) are I/O devices that enable human interaction with the computer system (keyboards, mice, touchscreens, game controllers)

I/O Communication Methods

• Programmed I/O (PIO) involves the CPU directly controlling the transfer of data between I/O devices and memory
  • The CPU continuously polls the I/O device to check for available data or completion of an operation
• Interrupt-driven I/O allows the CPU to perform other tasks while waiting for I/O operations to complete
  • The I/O device sends an interrupt signal to the CPU when it requires attention, allowing the CPU to handle the I/O operation
• Direct memory access (DMA) enables I/O devices to access main memory independently of the CPU
  • DMA controllers manage the transfer of data between I/O devices and memory, freeing up the CPU for other tasks
• Memory-mapped I/O (MMIO) uses a portion of the memory address space to represent I/O devices
  • The CPU reads from or writes to specific memory addresses to communicate with I/O devices
• Port-mapped I/O (PMIO) uses a separate address space, called the I/O address space, to communicate with I/O devices
  • The CPU uses special instructions (IN and OUT) to read from or write to I/O ports
• Polling involves the CPU periodically checking the status of an I/O device to determine if it is ready for a new operation or has completed a previous one
  • Polling can be inefficient if the I/O device is slow or not frequently used

I/O Performance Metrics

• Throughput measures the amount of data transferred per unit of time, typically expressed in bits per second (bps) or bytes per second (B/s)
  • Higher throughput indicates better performance
• Response time is the total time between the submission of an I/O request and its completion, including any delays due to processing, queuing, or transmission
• Latency is the time delay between the initiation of an I/O operation and its completion, often measured in milliseconds (ms) or nanoseconds (ns)
  • Lower latency indicates faster response times and better performance
• Bandwidth is the maximum amount of data that can be transferred per unit of time, typically measured in bits per second (bps) or bytes per second (B/s)
• I/O operations per second (IOPS) measures the number of input/output operations a device can perform in one second
  • Higher IOPS indicates better performance, especially for random access workloads
• Queue depth refers to the maximum number of outstanding I/O requests that a device can handle simultaneously
  • A higher queue depth allows for more concurrent I/O operations, potentially improving performance
• Seek time is the time required for a storage device (hard drive) to position its read/write head over the desired location on the disk
  • Lower seek times contribute to better overall I/O performance

I/O Software and Device Drivers

• Operating systems manage I/O devices and provide a consistent interface for applications to interact with them
• Device drivers are software components that enable the operating system to communicate with specific I/O devices
  • Device drivers abstract the hardware details and provide a standardized interface for the operating system
• I/O schedulers optimize the order and timing of I/O requests to improve performance and fairness
  • Examples of I/O schedulers include NOOP, Deadline, and Completely Fair Queuing (CFQ)
• Buffering is a technique used to temporarily store data in memory to reduce the frequency of I/O operations and improve performance
• Caching stores frequently accessed data in a faster storage medium (RAM) to reduce the need for slower I/O operations
• Spooling is a technique used to manage I/O requests for shared resources, such as printers
  • Print jobs are stored in a buffer (spool) and processed sequentially to avoid conflicts and optimize resource usage
• Error handling and recovery mechanisms are implemented in I/O software to detect and handle I/O errors gracefully (timeouts, retries, error codes)

Advanced I/O Techniques

• Non-volatile memory express (NVMe) is a high-performance interface for connecting solid-state drives (SSDs) to a computer system over PCIe
  • NVMe offers lower latency and higher throughput compared to traditional SATA interfaces
• Remote direct memory access (RDMA) allows direct memory access between two computers over a network without involving the CPU
  • RDMA reduces latency and CPU overhead in network I/O operations
• Solid-state drives (SSDs) use non-volatile memory (NAND flash) to store data, offering faster read/write speeds and lower latency compared to traditional hard disk drives (HDDs)
• Non-uniform memory access (NUMA) architectures have multiple memory nodes, each associated with a subset of the system's processors
  • NUMA-aware I/O scheduling can improve performance by prioritizing I/O operations that access local memory nodes
• Quality of Service (QoS) mechanisms prioritize I/O requests based on predefined policies or application requirements
  • QoS helps ensure that critical I/O operations receive the necessary resources and performance
• Virtualization technologies, such as virtual machines (VMs) and containers, introduce additional layers of I/O abstraction and management
  • Virtual I/O devices and virtual I/O schedulers optimize I/O performance in virtualized environments

Real-World Applications

• Databases rely heavily on I/O performance for data storage and retrieval
  • Optimizing I/O operations (indexing, caching, partitioning) is crucial for database performance
• File systems manage the storage and retrieval of files on storage devices
  • I/O performance directly impacts file system responsiveness and throughput
• Streaming services (video, audio) require high-throughput I/O to deliver content smoothly to users
  • Efficient I/O management ensures uninterrupted playback and reduces buffering
• Big data processing involves handling large volumes of data, often requiring distributed I/O across multiple nodes
  • Optimizing I/O performance is essential for timely data ingestion, processing, and analysis
• Cloud computing relies on efficient I/O performance to support various services (storage, databases, content delivery)
  • I/O virtualization and optimization techniques are employed to ensure consistent performance across multiple tenants
• Gaming applications demand high-performance I/O for real-time rendering, asset loading, and user input processing
  • Low-latency I/O is critical for responsive gameplay and immersive experiences
• Internet of Things (IoT) devices generate and consume data through various I/O interfaces (sensors, actuators, network)
  • Efficient I/O management is necessary to handle the scale and diversity of IoT data flows