What is Batch?
Batch is Google’s fully managed service for running batch workloads. The service handles resource provisioning, job scheduling and cleanup automatically. You define jobs as containers, Batch runs them on optimal infrastructure.
Core Features
- Automatic Provisioning: Compute resources are created on demand and deleted after job completion
- GPU and TPU Support: Native support for NVIDIA GPUs and Cloud TPU for ML and rendering
- Job Dependencies: Definition of task dependencies for complex workflows
- Spot VMs: Use of spot instances for up to 91% cost savings
- Array Jobs: Parallel execution of thousands of similar tasks
Typical Use Cases
High Performance Computing
Scientific simulations, financial modeling and engineering calculations benefit from massive scaling and GPU support.
Video Transcoding
Batch processing of large video libraries for various formats and resolutions. Each video is processed as a separate task.
3D Rendering
Rendering animations and visual effects distributed across hundreds of GPUs. Frames are rendered in parallel and stored in Cloud Storage.
Benefits
- No cluster administration required
- Pay only for actually used resources
- Massive scaling to thousands of CPUs or GPUs
- Integration with Cloud Storage and other GCP services
Integration with innFactory
As a Google Cloud Partner, innFactory supports you with Batch: workload analysis, job design, cost optimization through Spot VMs and integration with existing HPC workflows.
Available Tiers & Options
Batch
- No cluster management
- Automatic scaling
- GPU and TPU support
- Only suitable for batch workloads
Typical Use Cases
Technical Specifications
Frequently Asked Questions
What is Batch?
Batch is a fully managed service for scheduling and running batch jobs. The service automatically provisions compute resources, runs jobs and cleans up afterwards.
How does Batch differ from Dataflow?
Batch is optimized for container-based batch jobs while Dataflow is designed for data pipelines with Apache Beam. Batch is better suited for HPC and rendering, Dataflow for ETL and streaming.
Which accelerators are supported?
Batch supports NVIDIA GPUs (T4, V100, A100, L4) and Cloud TPU. Accelerators can be configured per task or per job.
How are jobs defined?
Jobs are defined as JSON or YAML with container image, resource requirements and dependencies. The API and gcloud CLI enable programmatic submission.
