Search and metadata in HLD - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Architecture

intermediate

Designing a scalable search system for metadata-rich documents

You need to design a search system that indexes documents with rich metadata (tags, authors, dates). Which architecture best supports fast, scalable search with frequent metadata updates?

A Use a distributed search engine like Elasticsearch with separate indices for metadata and content.

B Use a relational database with full-text search and update metadata in place.

C Store documents in a NoSQL key-value store and perform search by scanning all keys.

D Use a file system with metadata stored in separate JSON files and search by reading files sequentially.

💡 Hint

Think about systems optimized for search and handling frequent updates efficiently.

Distributed search engines like Elasticsearch are designed for fast, scalable search and can handle frequent metadata updates by indexing separately. Relational databases are less efficient for large-scale full-text search. Scanning all keys or files sequentially is slow and not scalable.

❓ scaling

intermediate

Estimating capacity for a metadata search service

Your metadata search service expects 10 million documents with an average of 20 metadata fields each. You expect 1000 queries per second. What is the best way to estimate the required hardware capacity?

A Estimate index size for metadata, average query complexity, and use benchmarks of similar systems to size CPU and memory.

B Calculate storage size for all metadata and content, then multiply by query rate to get CPU needs.

C Count number of documents only and assign one CPU core per 1000 documents.

D Ignore metadata size and focus only on network bandwidth for query traffic.

💡 Hint

Consider both data size and query complexity, and use real-world benchmarks.

Estimating capacity requires understanding index size, query complexity, and benchmarking similar systems. Simple document counts or ignoring metadata size leads to inaccurate estimates.

❓ tradeoff

advanced

Choosing between inverted index and graph database for metadata search

You must choose a data structure for searching documents by metadata. Which option best fits a use case with complex relationships between metadata (e.g., authors collaborating, hierarchical tags)?

A Use a relational database with multiple join tables for metadata relationships.

B Use an inverted index optimized for keyword search across metadata fields.

C Use a simple key-value store with metadata serialized as JSON strings.

D Use a graph database to model and query complex relationships between metadata entities.

💡 Hint

Consider which data structure naturally represents relationships and supports complex queries.

Graph databases are designed to model and query complex relationships efficiently. Inverted indexes excel at keyword search but not relationship queries. Relational joins can be expensive at scale. Key-value stores lack query flexibility.

🧠 Conceptual

advanced

Understanding metadata freshness impact on search results

In a search system with frequent metadata updates, what is the main tradeoff when choosing between real-time indexing and batch indexing?

A Real-time indexing reduces query latency but increases network bandwidth; batch indexing reduces bandwidth usage.

B Batch indexing improves freshness but requires more storage; real-time indexing reduces storage needs.

C Real-time indexing improves freshness but increases system load; batch indexing reduces load but delays updates.

D Batch indexing allows partial updates; real-time indexing requires full reindexing each time.

💡 Hint

Think about how update frequency affects system performance and data freshness.

Real-time indexing keeps data fresh but adds load due to constant updates. Batch indexing reduces load by updating less often but causes delays in reflecting changes.

❓ component

expert

Request flow in a distributed metadata search system

Trace the request flow when a user searches for documents by metadata in a distributed search system with multiple shards and a metadata cache layer. Which sequence correctly describes the flow?

A 1, 3, 4, 2, 5, 6

B 1, 2, 3, 4, 5, 6

C 1, 2, 4, 3, 5, 6

D 1, 3, 2, 4, 5, 6

💡 Hint

Consider cache lookup before querying shards and updating cache after aggregation.

The correct flow starts with user query to API gateway, cache check, forwarding to shards on miss, shards returning partial results, aggregation, returning results, and cache update.