Performance

[nasmart]

Please share your throughput 🚀

I'm interested to see what throughput users of this method of transferring data from BigQuery to AlloyDB are seeing.

Some of my results are shared below.

london_bicycles.cycle_hire

BigQuery table details

• Number of rows: 83,434,866
• Total logical bytes: 9.57 GB
• AlloyDB vCPU: 32
• Worker vCPU: 32
• Workers: 32

Throughput

• 1.5 TB/h (based on BigQuery data volume)
• 1.8 TB/h (based on AlloyDB data volume)

Private dataset table

BigQuery table details

• Number of rows: 23,390,740,736
• Number of partitions: 64
• Total logical bytes: 2.94 TB
• AlloyDB vCPU: 64
• Worker vCPU: 64
• Workers: 64

Throughput

• 2.7 TB/h (based on BigQuery data volume)
• 3.1 TB/h (based on AlloyDB data volume)

In the test cases above monitoring host CPU utilisation showed that the CPU available to the Java process is way over what's required (I didn't want this to be a bottleneck).

Please share your results.