Joschka Braun

Via `aws ssm get-parameters --names /aws/service/ecs/optimized-ami/amazon-linux-2/arm64/recommended` you can find the amiId of the recommended AMI for Arm64-based CPU and put it into your _spotty.yaml_ file.

@jemmyshin is working on an executor which will be useful for us. He will open a PR for it next week. I will follow-up & pair with him then.

Are we allowed to create low-resolution copies of the data..?

Make Jina now usable only from command line.

Increasing `cpu`, `mem` or `request_size` doesn't improve latency: ```bash Benchmarking with mem=4G, cpu=4.0 ╭────────────────── 🎉 Flow is available! ──────────────────╮ │ │ │ ID eee7e5b9fc │ │ Endpoint(s) grpcs://nowtest-eee7e5b9fc.wolf.jina.ai │ │...

Using `autoscaling` from jcloud: Initial results are worse but more are pending. If they continue to be worse: pairing with WOLF + manually increasing replicas.

Autoscaling doesn't work to well in current set-up, but manually updating `cpu` and `replicas` gives nice speed up from around 5 minutes with `cpu=0.1` and `replicas=1` to 2 minutes with...

Data Indexed vs. Query Time (s) for `DocarrayIndexerV2`:  Data Indexed vs. Query Time (s) for `AnnLiteIndexer` with 4gb RAM and 0.1 CPU for `n_dim=512`:  Data Indexed vs. Query...

Observed the following stability of AnnLite for different cpu configurations when indexing up to 1 min. dataset. Based on that will resources be allocated   

https://jina-ai.slack.com/archives/C0169V26ATY/p1657616879044809?thread_ts=1657574281.102059&cid=C0169V26ATY