Delete then Write vs. Read and Filter

[steff_pistachio]

Hello 👋
The docs suggest doing a delete then write if you want to maintain a single value in a column: https://cloud.google.com/bigtable/docs/keep-only-latest-value

I'm wondering if it's generally more performant to do it this way, or if it would be better to have a max version = 1 policy on the column family, and then when reading, apply a filter to get the most recent version of the data since it would return multiple values until garbage collection occurs.

Thanks

[dmmcerlean · Accepted solution]

Hi there,

The recommended approach (eagerly deleting old data when writing) is the most performant. It frees up storage faster, which reduces the amount of data reads need to sort through. Note that the recommendation is to delete and then rewrite the data in one atomic request,  not two separate requests.