Databases
Cloud Bigtable
Fully managed wide-column NoSQL — for massive analytical and operational workloads
AWS equivalent
DynamoDB (for scale) / No direct equivalent for time-series
AWS → GCP: Key Differences
- ▸
Bigtable is the technology behind Google Search, Maps, Gmail indexing — built for petabyte-scale.
- ▸
HBase API compatible: if you run HBase on-prem, migrating to Bigtable requires minimal code changes.
- ▸
Not fully serverless — you provision nodes. But nodes can be added/removed dynamically.
Key Concepts to Know
- 1
Row key design is critical — data is sorted by row key. Bad row key = hotspotting = poor performance.
- 2
No secondary indexes — design your row key to support your access patterns.
- 3
HBase compatible API: use existing HBase code with minimal changes.
- 4
Use cases: time-series (IoT, monitoring), recommendation systems, fraud detection, ad tech, financial data.
- 5
SLA: 99.99% for multi-cluster, 99.9% for single-cluster.
DCE Interview Tips
- ★
Bigtable is the right answer for: IoT sensor data, stock price history, clickstream data, anything where you write once and read by time range.
- ★
'In your Thai factory IoT scenario — thousands of sensors writing millions of data points per minute — Bigtable handles that at single-digit millisecond latency.'
- ★
Bigtable vs BigQuery: 'Bigtable for operational, low-latency reads/writes. BigQuery for analytical SQL queries across historical data.'
Common Gotchas
- !
Minimum recommended: 3 nodes for production. Each node costs ~$0.65/hour.
- !
No SQL support — it's a NoSQL key-value store.
- !
Bigtable doesn't support multi-row transactions like a relational DB.