Examkin

Question

Difficulty: MediumPlanning Database Options and Persistent Disk Types

A financial analytics enterprise is architecting a new microservice on Google Cloud to ingest and analyze real-time credit card transaction telemetry for fraud detection. The workload requires ingesting millions of event writes per second with single-digit millisecond latency, supporting time-series data access patterns, and scaling horizontally without schema migration overhead. High-level relational features such as multi-table ACID transactions and complex SQL joins are not required. Which database service should be selected to fulfill these requirements with minimal operational management?

  1. Cloud BigtableAnswer
  2. B
    Cloud SQL for PostgreSQL configured with read replicas
  3. C
    Cloud Spanner configured in a multi-region deployment
  4. D
    A self-managed Cassandra cluster deployed on Spot Compute Engine instances

Answer

Cloud Bigtable should be selected because it is designed for high-throughput time-series and key-value workloads with low latency and seamless horizontal scaling.
Cloud Bigtable is a fully managed NoSQL wide-column database service ideal for high-volume telemetry, time-series data, and key-value stores. It delivers consistent single-digit millisecond latency and scales linearly to millions of write operations per second with low operational overhead.

Step-by-Step Solution

1
Analyze the workload requirements
Identified key constraints: millions of writes per second, single-digit millisecond latency, time-series data patterns, NoSQL key-value access, and minimal operational overhead.
Matching workload access patterns to Google Cloud database services ensures optimal performance and cost efficiency.
2
Evaluate database candidates against latency and write throughput needs
Cloud Bigtable provides scalable NoSQL wide-column storage optimized specifically for high-throughput ingestion and time-series data.
Relational options like Cloud SQL cannot handle millions of writes per second, and Cloud Spanner is intended for relational transactional workloads rather than simple time-series key-value data.
3
Assess operational viability of self-managed options on Spot instances
Self-managed clusters on Spot instances violate Google Cloud operational best practices for stateful applications due to instance preemptibility.
Managed services minimize administrative overhead while providing necessary SLA guarantees.

Key Concept

Selecting Cloud Bigtable for high-throughput NoSQL time-series and key-value workloads
Rate this question

Topics

Planning Database Options and Persistent Disk Types

More questions from this topic