11.9.1: Review Questions

Review Questions

Multiple Choice

1. Which cloud storage type offers the highest scalability and lowest cost per GB but has higher access latency than alternatives?
   • a) Block storage
   • b) File storage
   • c) Object storage
   • d) Database storage
2. In HDFS, what is the default number of replicas maintained for each data block?
   • a) 1
   • b) 2
   • c) 3
   • d) 5
3. Google Spanner is an example of which database category?
   • a) NoSQL (document store)
   • b) Traditional relational (single-node SQL)
   • c) NewSQL (globally distributed SQL with strong consistency)
   • d) Key-value store

Short Answer

4. Compare object storage, block storage, and file storage across four dimensions: access method, scalability, latency, and ideal use case. Create a comparison table.
5. Explain the difference between strong consistency and eventual consistency. Why might a social media platform choose eventual consistency for its news feed, while an online banking application requires strong consistency?
6. What is the role of the NameNode in HDFS? Why is it a potential single point of failure, and how do modern HDFS deployments address this vulnerability?
7. Describe a cloud backup strategy using the 3-2-1 rule (3 copies, 2 different media, 1 offsite). How would you implement this using cloud services? What are RPO and RTO, and how do they influence your backup design?

True/False

8. True or False: Object storage supports in-place modification of data — you can update a portion of an object without rewriting the entire object.
9. True or False: Netflix uses Amazon S3 as the primary backup store for all its data, ensuring any corruption can be mitigated by restoring from S3.
10. True or False: NoSQL databases always sacrifice consistency for availability and can never provide ACID transactions.

Answer Key:

1. c
2. c
3. c
4. Table: Object (API/HTTP, virtually unlimited, higher latency, backups/media/data lakes); Block (mounted volume, limited per volume, lowest latency, databases/OS disks); File (NFS/SMB path, moderate, moderate latency, shared files/legacy apps).
5. Strong: all reads see latest write instantly (banking — account balance must be accurate). Eventual: replicas may temporarily differ (social media — seeing a "like" 2 seconds late is acceptable). Trade-off is performance/availability vs. correctness guarantees.
6. NameNode maintains directory tree, file metadata, and block-to-node mappings. Single point of failure because losing it means losing the file system index. Modern solutions: standby NameNode (automatic failover), HDFS Federation (multiple NameNodes for different namespaces).
7. 3-2-1 in cloud: primary data in managed DB (copy 1), automated snapshots to same-region storage (copy 2, different media), cross-region replication to another region (copy 3, offsite). RPO determines backup frequency; RTO determines recovery architecture (cold standby vs. hot failover).
8. False
9. True
10. False (many NoSQL databases now support ACID transactions, e.g., MongoDB, DynamoDB)