Time Required to embed a column in AlloyDB using in-built text-embedding methods

Google Cloud Database
1

I am not sure how long it takes to embed or vectorize a column in AlloyDB. I am using in-built google-ml-integration with text-geckoembedding option for embedding() function. The query timed out on AlloyDB studio and from psql shell on a CE. Is there an alternate or better approach to embedd data in AlloyDB?

Isn’t AlloyDB claimed to be doing this things faster as it was built with additional capabilities for embedding and AI stuff.

I even gave a try to run batches but it still timedout. I am planning to use local machine GPUs to process and then store the results in db.

2

I noticed that when you execute the Update query on PSQL for embeddings over Compute Engine, the embeddings don’t appear immediately in AlloyDB Studio even if the CE session terminates. However, they do show up the next day when you run a SELECT query on the same table. This method has consistently worked for me generating and viewing embeddings on entire columns having 400K+ records.