Block Alignment and Unaligned I/O Overhead

Unaligned I/O Costs Extra Work

Operating systems and devices prefer I/O sizes and offsets that are multiples of the block size (often 4096). An unaligned range may require reading two blocks, modifying the middle, and writing both back even if the application wanted a few bytes.

Alignment Arithmetic

Align start down to block boundary, end up to boundary, compute span in blocks.

c
Loading…

O_DIRECT bypasses page cache and enforces alignment strictly
Databases pad headers to page boundaries for mmap
SSD erase blocks are larger than OS pages; alignment still helps FTL
Measure read amplification on misaligned random updates

Working Through the Exercise

Keep the relevant documentation open while you implement. When your output disagrees with the reference, trace one failing case by hand before changing random lines.

Trace first: walk one failing input step by step
Read the manual: skim the official doc for the mechanism named in this task
Compare baselines: save measurements from the prior step so speedups are honest

Why for this exercise

You will compute how many blocks an unaligned range touches versus an aligned range. This exercise asks you to print overhead for sub-page updates.

Write a C program that demonstrates the cost of unaligned I/O.

Requirements:

Define a block size of 4096 bytes
Simulate writing 8192 bytes (2 blocks) under three scenarios:
- Aligned: offset=0, size=8192 — 2 full block writes
- Offset-unaligned: offset=1024, size=8192 — spans 3 blocks, first and last are partial
- Size-unaligned: offset=0, size=9000 — spans 3 blocks, last is partial
For each scenario, calculate:
- Number of physical blocks touched
- Number of full block writes
- Number of read-modify-write operations
- Total bytes read from disk (for RMW)
- Total bytes written to disk
- Write amplification factor: total bytes written / user bytes requested
Print a comparison table
Include a note about Direct I/O (O_DIRECT) and why it requires aligned buffers
Explain why databases use posix_memalign for I/O buffers

Unaligned I/O Costs Extra Work

Alignment Arithmetic

Align start down to block boundary, end up to boundary, compute span in blocks.

c
Loading…

O_DIRECT bypasses page cache and enforces alignment strictly
Databases pad headers to page boundaries for mmap
SSD erase blocks are larger than OS pages; alignment still helps FTL
Measure read amplification on misaligned random updates

Working Through the Exercise

Keep the relevant documentation open while you implement. When your output disagrees with the reference, trace one failing case by hand before changing random lines.

Trace first: walk one failing input step by step
Read the manual: skim the official doc for the mechanism named in this task
Compare baselines: save measurements from the prior step so speedups are honest

Why for this exercise

You will compute how many blocks an unaligned range touches versus an aligned range. This exercise asks you to print overhead for sub-page updates.

Write a C program that demonstrates the cost of unaligned I/O.

Requirements:

Define a block size of 4096 bytes
Simulate writing 8192 bytes (2 blocks) under three scenarios:
- Aligned: offset=0, size=8192 — 2 full block writes
- Offset-unaligned: offset=1024, size=8192 — spans 3 blocks, first and last are partial
- Size-unaligned: offset=0, size=9000 — spans 3 blocks, last is partial
For each scenario, calculate:
- Number of physical blocks touched
- Number of full block writes
- Number of read-modify-write operations
- Total bytes read from disk (for RMW)
- Total bytes written to disk
- Write amplification factor: total bytes written / user bytes requested
Print a comparison table
Include a note about Direct I/O (O_DIRECT) and why it requires aligned buffers
Explain why databases use posix_memalign for I/O buffers

Topic Overview

Unaligned I/O Costs Extra Work

Alignment Arithmetic

Working Through the Exercise

Why for this exercise

Implementation Task

Progressive Hints

Resources

Open Source

Block Alignment and Unaligned I/O Overhead

Topic Overview

Unaligned I/O Costs Extra Work

Alignment Arithmetic

Working Through the Exercise

Why for this exercise

Implementation Task

Progressive Hints

Resources

Open Source