Overview
This lecture explains the components and structure of Google BigQuery pricing, highlighting storage, compute, and other cost factors to help users avoid unexpected charges.
BigQuery Pricing Components
- BigQuery charges primarily for storage, compute (query processing), streaming inserts, BI Engine memory, and data transfers/exports.
- Free tiers include 10 GB storage and 1 TB query data per month.
Storage Pricing
- Storage is split into active (accessed within 90 days) and long-term (not accessed for 90 days) categories.
- Active storage costs $0.02/GB/month; long-term storage is $0.01/GB/month (a 50% reduction).
- Storage billing models: Logical (uncompressed data size, includes metadata) and Physical (compressed data size).
- Logical storage typically costs less per GB than physical, but actual charges depend on data compression and usage.
- The storage billing model can be set at the dataset level; changing the model takes about 24 hours to update.
Compute (Query) Pricing
- Compute charges are based on the amount of data processed, not the number of queries or rows returned.
- Using SELECT * scans all columns and costs more than selecting specific columns.
- Two pricing models:
- On-demand: $6.25 per TB processed (first 1 TB per month is free), suitable for variable workloads.
- Flat-rate: Pre-purchase dedicated slots for a fixed monthly fee, ideal for predictable/high workloads.
- Costs vary by region; always check the current rates for your region.
Additional Costs and Free Operations
- Data transfers, streaming inserts ($0.0134 per 200 MB), and cross-region replication have separate charges.
- Batch data loads, batch exports, and metadata operations are free.
- BigQuery ML model creation is charged at a higher rate (e.g., $400 per TB for regression).
- Moving data between regions can incur additional costs depending on locations involved.
Key Terms & Definitions
- Active Storage — Data accessed within the last 90 days, billed at the regular rate.
- Long-term Storage — Data not accessed for 90 days, billed at a reduced rate.
- Logical Storage — Billing based on uncompressed data size.
- Physical Storage — Billing based on the compressed, stored data size.
- Slot — Computational unit used to process queries in BigQuery’s flat-rate pricing model.
Action Items / Next Steps
- Review and choose appropriate storage and compute billing models for your datasets and workload.
- Analyze query construction to minimize costs (avoid SELECT *).
- Check official BigQuery pricing page for up-to-date rates and region-specific costs.
- If unsure, consult with Google Cloud or an expert to optimize your billing model.