CSE Roadmap

Core tier · about 3 hours · 6 parts

Part 1 of 6 · 1 min read

Why this page comes sixth

In this module you will

  • Explain striping, mirroring and parity, and say what each gives and costs.
  • Compute the usable capacity and the failures that RAID 0, 1, 5, 6 and 10 survive.
  • Compute parity with XOR, rebuild a lost block, and update parity after a small write.
  • Estimate the random write rate of an array and explain the small-write penalty of RAID 5.
  • Say what RAID does not do, and why it is not a backup.

Pages 1 to 5 dealt with one device. One device has a limited speed and it can fail. This page combines several devices into one faster or safer one. The idea of keeping redundant data so that a failure can be survived returns in Level 9 (file system checks and journals) and in Level 12 (production storage). The parity arithmetic here is simple and is a good first example of an error-correcting code.

About this example About this example

Real: The C program in section 3 was compiled with gcc -Wall -Wextra and run on a Linux test machine (2 processors, kernel 6.18). Its output was checked by hand for one parity byte. The test machine has no software RAID, and /proc/mdstat does not exist on it.

Scripted: The array of six 4 TB disks, 150 random operations per second for each disk, the 100 MB/s rebuild rate and the 2% yearly failure rate are made up to show the idea.

Not real: The model ignores controller caches, write-back batching, hot spares, and failures that happen together because disks of one batch age alike. Real arrays do better on some loads and worse on others.