Storage 1: OS input/output, memory hierarchy

Overview

Full lecture notes on storage — Textbook readings

System calls for reading and writing

• Process isolation means processes don’t directly interact
• They can interact via explicit, checked input/output (I/O) system calls
• Some early operating systems offered many I/O system calls for different kinds of device
  • read_tape, write_tape, read_screen, write_screen, read_record, write_record…
• Modern operating systems unify devices into a minimal set of system calls

File descriptor system calls

• ssize_t read(int fd, void* buf, size_t sz)
• ssize_t write(int fd, const void* buf, size_t sz)
• Transfer sz bytes of data between an abstract device fd and a memory buffer buf
  • read: Read bytes from device to memory
  • write: Write bytes from memory to device
• Abstract devices are represented in system calls with file descriptors fd of type int

Creating file descriptors

• Processes start with 3 standard file descriptors
  • STDIN_FILENO = 0: standard input (stdin, std::cin)
  • STDOUT_FILENO = 1: standard output (stdout, std::cout)
  • STDERR_FILENO = 2: standard error (stderr, std::cerr)
• Processes can create new file descriptors
• int open(const char* filename, int oflag, ...)
  • filename names the device or sub-device to open
  • oflag is O_RDONLY, O_WRONLY, O_WRONLY | O_CREAT, …
  • Returns a previously-unused file descriptor
  • int close(int fd) closes an open file descriptor
• Many other system calls create fds for different kinds of device, including pipe, socket, …

Devices and files

• Some filenames correspond to whole devices
  • Example: /dev/nvme2n1, /dev/console
• But whole devices are enormous and inconvenient
  • /dev/nvme2n1 is 75GB!
• The kernel should offer useful abstractions that simplify process programming
• File systems break enormous disks into named sections called files
  • Can create, extend, overwrite, rename
  • Kernel file system implementation maps files to device storage

Losing power

• Primary memory does not survive power outages or reboots
• write(fd, buf, sz) moves data from primary memory buf to some abstract device fd
• What should happen on power outage or reboot?

Durable storage

• Storage is durable or persistent when modifications survive power outage or reboot
  • Volatile storage is the opposite of durable storage
• Durable storage is inherently slower than volatile storage

Creating a file one byte at a time

• w-syncbyte: Create a file one byte at a time, durably
  • The OS guarantees that every byte is written durably
• w-osbyte: Create a file one byte at a time, non-durably
  • The OS does not guarantee that every byte is written durably
  • “A successful return from write() does not make any guarantee that data has been committed to disk.”
  • The OS will write the bytes durably, eventually (within 30 seconds, but usually much sooner)
• w-stdiobyte: Create a file one byte at a time, non-durably, using a library (stdio)

Write a file one byte at a time using system calls

    size_t n = 0;
    while (n < size) {
        ssize_t r = write(fd, buf, 1);
        if (r != 1) {
            perror("write");
            exit(1);
        }
        ++n;
    }

Write a file one byte at a time using the stdio library

    size_t n = 0;
    while (n < size) {
        int ch = fputc(buf[0], f);
        if (ch == EOF) {
            perror("write");
            exit(1);
        }
        ++n;
    }

• fputc(buf[0], f) ≈ fwrite(buf, 1, 1, f), with different error return convention
  • Check the manual!

Question 1

• How long will it take each program to write data one byte at a time?
• Will the results differ by operating system? How much?

Investigating standard I/O

• strace
  • Powerful Linux program snoops on another program’s system calls
  • Prints a human-readable summary of those system calls to a file
• strace -o strace.out PROGRAM ARGUMENTS…
  • -f: follow forks (subprocesses)
  • -s N: Print more bytes of string arguments/return values
  • -e trace=open,openat,close,read,write: Only print those system calls
  • man strace for more

What is standard I/O doing?

Cache

• A cache is an area of fast storage used to speed up access to underlying storage
• Computer systems are rife with caches
  • One of the most important patterns you learn in this class
  • Registers as a cache for primary memory
  • Memory as a cache for flash/hard disk
• Life is rife with caches