--- title: "Single master not a GFS bottleneck" subject: "Big Data Technologies" chapter: 2 key: "semester_8/big_data_technologies:gfs-single-master-bottleneck" source: https://www.bctengineer.com/q/semester_8/big_data_technologies/gfs-single-master-bottleneck site: BCT Engineering --- # Single master not a GFS bottleneck Big Data Technologies · Chapter 2: Google File System - Asked in: 2073 Bhadra, 5 marks - 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-single-master-bottleneck - Whole chapter: https://www.bctengineer.com/md/semester_8/big_data_technologies/chapter-2 ## Question Why single master is not a bottleneck in GFS cluster? ## Short answer The master handles only metadata, not data transfers. Large chunk size (64 MB) and client caching minimize requests. Metadata in memory gives fast access, and write leases offload coordination, so a single master scales to thousands of nodes. ## Answer GFS uses a single master to keep the design simple and to allow global decisions (chunk placement, replication). A single master could limit scalability, so GFS is designed to keep its involvement in each operation minimal. #### Reasons it is not a bottleneck 1. **Master is not in the data path:** clients ask the master only for metadata (chunk handle and locations). The actual reading and writing of data happens directly between clients and chunkservers. 2. **Large chunk size (64 MB):** fewer chunks per file means fewer metadata requests to the master. A client reading a large file contacts the master once per 64 MB. 3. **Client caching of metadata:** the client caches chunk locations for some time, so repeated accesses to the same chunk do not contact the master again. 4. **Batching:** a client can ask for several chunks in one request, and the master can reply with information about the chunks that follow, at almost no extra cost. 5. **All metadata is in memory:** the master serves requests quickly, and scanning memory for garbage collection and re-replication is efficient. Metadata is small (under 64 bytes per chunk). 6. **Light-weight operations:** the master does not persist chunk locations; it learns them from chunkservers. Only namespace and mapping changes go to the operation log, which is checkpointed (compact B-tree-like form) to keep restart fast. 7. **Leases:** the master delegates control of each write to a primary chunkserver by a lease, so it does not coordinate every mutation. 8. **Replicas and shadow masters:** the operation log is replicated, and shadow masters provide read-only service, reducing load and failure risk. Therefore a single master handles the metadata work of a cluster with thousands of chunkservers and hundreds of clients without becoming a performance limit. ## Check yourself The answer's key points as multiple-choice questions, with the correct option and why. 1. Why is the single master not a bottleneck for data transfer? A. All data passes through it very fast B. Clients get only metadata from it; data flows directly between clients and chunkservers C. It stores data on SSD D. It has many copies of data Answer: B. The master serves only metadata; actual reads and writes go directly to chunkservers. (Key point: Master not in data path) 2. How does client caching help reduce master load? A. Clients cache chunk locations so repeated access to a chunk needs no master contact B. Clients cache the whole file C. Clients replace the master D. Clients cache the operation log Answer: A. Cached chunk locations let clients skip the master for repeated accesses to the same chunk. (Key point: Client caching) 3. What is the role of leases in keeping the master lightly loaded? A. They lock the entire file system B. They delete old chunks C. They give clients write permission forever D. The master delegates write ordering to a primary chunkserver, so it does not coordinate every mutation Answer: D. A lease makes one replica primary to order mutations, taking that work off the master. (Key point: Leases) 4. Which is true about what the master writes to its operation log? A. Chunk locations are logged on every change B. File data is logged C. Only namespace and mapping changes are logged; chunk locations are learned from chunkservers D. Nothing is logged Answer: C. Chunk locations are not persisted; they are learned from chunkservers, keeping the log small and restarts fast. (Key point: Chunk locations not persisted) ## Related in Big Data Technologies - [GFS master role in read/write](https://www.bctengineer.com/question.md?key=semester_8%2Fbig_data_technologies%3Agfs-master-read-write-flow) — Big Data Technologies, chapter 2 ## Related in other subjects - [Single phase motor not self-starting; DFRT](https://www.bctengineer.com/question.md?key=semester_4%2Felectrical_machines%3Aspim-not-self-starting-dfrt) — Electrical Machines, chapter 7, asked 3× - [Why single-phase IM is not self starting; methods](https://www.bctengineer.com/question.md?key=semester_5%2Felectric_machines_ii%3Asingle-phase-not-self-starting-methods) — Electric Machines II, chapter 3, asked 2× - [Single phase motor not self-starting; two starting methods](https://www.bctengineer.com/question.md?key=semester_4%2Felectrical_machines%3Aspim-not-self-starting-starting-methods) — Electrical Machines, chapter 7 - [Reason: single phase motor not self-starting](https://www.bctengineer.com/question.md?key=semester_4%2Felectrical_machines%3Aspim-not-self-starting-reason) — Electrical Machines, chapter 7