[Feature] [BanyanDB] Support no SeriesID for relative small dataset of record

[wu-sheng]

Search before asking

• [X] I had searched in the issues and found no similar feature requirement.

Description

BanyanDB requires SeriesID for every Stream to shard and unique the record row. But in some cases, such as newly added SpanAttachedEvent, records are not generated on-demand and only in specific scope(for specific instance/pod in the example case). So the size of the dataset would not be huge, but we can't tell how big it would do, and we don't have a suitable field in the record entity as a SeriesID.

So, we should create a SeriesID mechanism, which generates ID on the server side. This ID should be generated in self-adaption. OAP could provide some configurations at the scale to assist.

Use case

We need to discuss more details about this when BanyanDB server picks this up.

Related issues

No response

Are you willing to submit a PR?

• [ ] Yes I am willing to submit a PR!

Code of Conduct

• [X] I agree to follow this project's Code of Conduct