A Dataflow pipeline in europe-west4 with max 3 n1-standard-1 workers reading Pub/Sub and writing to BigQuery EU is CPU-bound at peak. Which two changes increase pipeline throughput?

When a Dataflow pipeline is CPU-bound, it means the workers are spending too much time processing data and not enough time on I/O or other tasks. To increase throughput, you need more processing power. Increasing the maximum number of workers allows Dataflow to scale out, distributing the processing load across more machines. Using a larger machine type for Dataflow workers provides each worker with more CPU and memory resources, allowing it to process data faster. Moving the pipeline to a different region (us-central1) or adding intermediate buffering with Bigtable or Spanner would not directly address the CPU bottleneck; these might introduce latency or additional complexity without solving the core processing limitation.