Indexing is painfully slow on MacBook M1 with 32Gb RAM

Support
1

As per title using Docker (Colima) on MacBook Pro M1

docker run --name marqo -p 8882:8882 marqoai/marqo:latest

When indexing documents performance is painfully slow - 30 smallish documents take almost 2 minutes or so.

Is there anyway this can be tweaked? This is an OOB configuration with a decent RAM (16Gb) and CPU (4) allocated to Colima VM.

When I look at TOP of the Colima VM I can see the Marqo web server (unicorn) maxing out on CPU and using very little RAM (circa 13%).

2

Hello @musicmakingal . Thanks for raising this issue. 30 smallish documents taking almost 2 minutes sounds really slow. Could you provide some more details so that we can help investigate the slowness issue? Thanks a lot.

  • What embedding model do you use for your index?
  • Do you have images (urls) in your document?
  • How many documents do you index in one batch?
  • Are you using py-marqo client to index the documents. If so, could you please turn on the telemetry to collect some metrics about the ingestion of your docs? See Analysing Processing Time - Marqo Docs for details about capturing telemetry

Regards,
Yihan