Announced May 2025: Dataproc Serverless is now Google Cloud Serverless for Apache Spark
Google Cloud Serverless for Apache Spark
Focus on your code, not your infrastructure
Run your Apache Spark jobs easier on a customizable zero-ops platform, smarter with Gemini assistance, and faster with the performance of Lightning Engine.
Apache Spark is a trademark of The Apache Software Foundation.
Features
Industry-leading performance
Supercharge your jobs with Lightning Engine, our next-generation vectorized engine. Get over 4.3x faster performance and lower TCO for your serverless Spark workloads, automatically.
Zero-Ops with intelligent autoscaling
Eliminate cluster management with intelligent autoscaling. Resources scale up, and down automatically to perfectly match your job's needs, ensuring maximum performance, and cost-efficiency without paying for idle time.
AI-powered development
Accelerate your entire workflow. Write and debug PySpark, Scala, and Java code with Gemini Code Assist in BigQuery Studio and launch GPU-accelerated environments with pre-configured ML Runtimes.
Unified Spark and SQL experience
Eliminate context switching. Develop and run your workloads in a single environment like BigQuery Studio, seamlessly blending powerful SQL with the flexibility of PySpark in the same notebook.
Two tiers of performance
| Two tiers of performance | Tiers to match your specific needs, from standard batch processing to the most demanding, performance-critical jobs. |
|---|---|
| Tier | Best for |
Standard | Ideal for cost-effective batch processing, data transformations, and general-purpose Spark jobs.
|
Premium | For the most demanding workloads, offering maximum performance with Lightning Engine, AI/ML acceleration, and interactive capabilities.
|
Two tiers of performance
Tiers to match your specific needs, from standard batch processing to the most demanding, performance-critical jobs.
Standard
Ideal for cost-effective batch processing, data transformations, and general-purpose Spark jobs.
- General purpose Spark ETL
- Scheduled data pipelines
- Cost-sensitive batch jobs
Premium
For the most demanding workloads, offering maximum performance with Lightning Engine, AI/ML acceleration, and interactive capabilities.
- Performance-critical jobs powered by Lightning Engine for 4.3x boost
- Interactive data science and analysis
- GPU-accelerated AI and ML
- Complex, large-scale data processing
Common Uses
Interactive Data Science
Interactive Data Science
Empower data scientists to explore data and rapidly iterate on Spark ML models. Unify SQL and Spark in a single BigQuery Studio notebook, moving seamlessly from data exploration with SQL to model building with PySpark without ever managing infrastructure.
Tutorials, quickstarts, & labs
Interactive Data Science
Interactive Data Science
Empower data scientists to explore data and rapidly iterate on Spark ML models. Unify SQL and Spark in a single BigQuery Studio notebook, moving seamlessly from data exploration with SQL to model building with PySpark without ever managing infrastructure.
Pricing
| Transparent, value-driven pricing | Serverless for Apache Spark pricing is based on per-second usage of compute (DCUs), GPUs, and shuffle storage. | |
|---|---|---|
| Services and usage | Subscription type | Price (USD) |
Data Compute Unit (DCU) | Standard | Starting at $0.06 per hour |
Premium | Starting at $0.089 per hour | |
Shuffle storage | Standard | Starting at $0.04 per GB/month |
Premium | Starting at $0.1 per GB/month | |
Accelerator pricing | a100 40 GB | Starting at $3.52069 per hour |
a100 80 GB | Starting at $4.713696 per hour | |
L4 | Starting at $0.672048 per hour | |
View pricing details for Google Cloud Serverless for Apache Spark.
Transparent, value-driven pricing
Serverless for Apache Spark pricing is based on per-second usage of compute (DCUs), GPUs, and shuffle storage.
Data Compute Unit (DCU)
Standard
Starting at
$0.06
per hour
Premium
Starting at
$0.089
per hour
Shuffle storage
Standard
Starting at
$0.04
per GB/month
Premium
Starting at
$0.1
per GB/month
Accelerator pricing
a100 40 GB
Starting at
$3.52069
per hour
a100 80 GB
Starting at
$4.713696
per hour
L4
Starting at
$0.672048
per hour
View pricing details for Google Cloud Serverless for Apache Spark.
Business Case
Build your business case for Google Cloud Serverless for Apache Spark
The economic benefits of Google Cloud Dataproc and Serverless Spark versus alternative solutions
See how Serverless for Apache Spark delivers significant TCO savings and business value compared to on-prem and other cloud solutions.
Related Content
In the report:
Discover how Dataproc and Serverless for Apache Spark can deliver 18% to 60% cost savings compared to other cloud-based Spark alternatives.
Explore how Google Cloud Serverless for Apache Spark can provide 21% to 55% better price-performance than other serverless Spark offerings.
Learn how Dataproc and Google Cloud Serverless for Apache Spark simplify Spark deployments and help reduce operational complexity.
FAQ
When should I choose Serverless for Apache Spark versus Dataproc?
Choose Serverless for Apache Spark when you want to focus on your code and eliminate all infrastructure management. It's ideal for new Spark pipelines, interactive analysis, and jobs with unpredictable demand where speed and simplicity are the priority.
Do I need to install my own libraries like PyTorch or XGBoost?
The Premium tier is designed for AI/ML and comes with pre-configured ML Runtimes that have common libraries like PyTorch, XGBoost, and scikit-learn built-in. This eliminates complex setup and allows you to get started with your data science workloads in minutes.
How do I get the best performance and how does pricing work?
For maximum performance, you can select the Premium tier, which is powered by Lightning Engine. Pricing is based on a "pay-for-what-you-use" model, where you are billed per second only for the duration of your job's execution. This is highly cost-effective as it eliminates the cost of idle clusters.