---
title: "GFS vs other file systems"
subject: "Big Data Technologies"
chapter: 2
key: "semester_8/big_data_technologies:gfs-differ-other-fs"
source: https://www.bctengineer.com/q/semester_8/big_data_technologies/gfs-differ-other-fs
site: BCT Engineering
---

# GFS vs other file systems

Big Data Technologies · Chapter 2: Google File System

- Asked in: 2075 Bhadra, 5 marks
- Video: [Google File System - Paper that inspired Hadoop](https://www.youtube.com/watch?v=eRgFNW4QFDc) — Defog Tech
- Answer checked by Claude Opus 5.5 (high effort) on 2026-10-14
- Page: https://www.bctengineer.com/q/semester_8/big_data_technologies/gfs-differ-other-fs
- Whole chapter: https://www.bctengineer.com/md/semester_8/big_data_technologies/chapter-2

## Question

How GFS differ from other File Systems? List out five distinct differences.

## Short answer

GFS differs by expecting failures on commodity hardware, using 64 MB chunks, favoring large streaming reads and append writes, separating a metadata master from direct client-to-chunkserver data flow, and providing relaxed consistency with replication and atomic record append rather than strong POSIX semantics.

## Answer

GFS was designed around Google's workload and failure assumptions, so it differs from traditional file systems (such as NFS, ext, NTFS or AFS) in these ways.

| Point | GFS | Traditional file systems |
|---|---|---|
| Hardware | Thousands of cheap commodity machines; failures are normal | Reliable, costly hardware; failure is rare |
| Chunk/block size | Very large, 64 MB chunks | Small blocks (4 KB to 64 KB) |
| File size and workload | Multi-GB files, large streaming reads, append-mostly writes | Many small files, random read/write |
| Architecture | Single master for metadata, chunkservers for data, clients talk directly to chunkservers | Often tightly coupled server (NFS) or no separate metadata service |
| Fault tolerance | Each chunk replicated (3 copies), checksums, automatic re-replication | Relies on RAID or backups, limited replication |
| Consistency | Relaxed consistency, atomic record append | Strong POSIX semantics |
| Caching | No client data cache (files too large); metadata cached | Data caching is common |
| Interface | Custom API with snapshot and record append, not fully POSIX | Standard POSIX interface |

Five distinct differences in brief: (1) commodity hardware with constant failures expected, (2) 64 MB chunks, (3) append-oriented large-file workload, (4) separate metadata master with direct client-chunkserver data flow, (5) built-in replication with a relaxed consistency model and atomic record append.

## Check yourself

The answer's key points as multiple-choice questions, with the correct option and why.

1. How does GFS's hardware assumption differ from traditional file systems?
   A. It assumes reliable costly hardware
   B. It uses commodity machines where failures are normal
   C. It needs RAID only
   D. It assumes no failures
   Answer: B. GFS runs on cheap commodity machines and treats failure as the normal case. (Key point: Hardware assumption)

2. Which block size does GFS use compared to traditional file systems?
   A. 4 KB to 64 KB in GFS
   B. 64 MB in GFS and 4 KB to 64 KB in traditional systems
   C. 64 KB in GFS and 64 MB in traditional systems
   D. Same size in both
   Answer: B. GFS uses very large 64 MB chunks to suit multi-GB files, reducing metadata. (Key point: Chunk size)

3. Which describes GFS consistency?
   A. Strict POSIX semantics
   B. No consistency at all
   C. Relaxed consistency with atomic record append
   D. Strong consistency with client data caching
   Answer: C. GFS relaxes consistency and offers atomic record append, unlike strong POSIX semantics. (Key point: Consistency model)

## Related in Big Data Technologies

- [Google File System architecture](https://www.bctengineer.com/question.md?key=semester_8%2Fbig_data_technologies%3Agfs-architecture) — Big Data Technologies, chapter 2, asked 2×
- [Availability and fault tolerance in GFS](https://www.bctengineer.com/question.md?key=semester_8%2Fbig_data_technologies%3Agfs-availability-fault-tolerance) — Big Data Technologies, chapter 2
- [Column vs row-oriented NoSQL databases](https://www.bctengineer.com/question.md?key=semester_8%2Fbig_data_technologies%3Acolumn-oriented-nosql-row-oriented) — Big Data Technologies, chapter 4

## Related in other subjects

- [GFS vs HDFS architecture](https://www.bctengineer.com/question.md?key=semester_8%2Finformation_system%3Agfs-vs-hdfs) — Information System, chapter 8
- [Microwave vs low frequency and five system protocols](https://www.bctengineer.com/question.md?key=semester_7%2Frf_and_microwave_engineering%3Amicrowave-differences-five-protocols) — RF and Microwave Engineering, chapter 1
- [File attributes; directory systems; free blocks](https://www.bctengineer.com/question.md?key=semester_6%2Foperating_system%3Afile-attribute-directory-free-blocks) — Operating System, chapter 5
- [File system layout; interface vs implementation](https://www.bctengineer.com/question.md?key=semester_6%2Foperating_system%3Afs-layout-interface-vs-implementation) — Operating System, chapter 5