Buckets:: segmentation per host/group or not at all?

InfluxDB 2
1

Hi! what is intended purpose of a bucket? is it like a database or like a table in a db?
Is there a performance optimization to have all metrics in one or could have to have multiple buckets?
I’m trying to transition from ganglia, well each metric had it’s own file, and all host metrics were in a host directory (which were kept in a group directory)
So, with this context, how should i organize buckets? (< 200 hosts, basic linux monitoring, so few hundredths metrics per host) Should i have per host buckets?
And if yes, is there a possibility to NOT have to pre-create buckets, and just configure telegraf to send to bucket = $HOSTNAME?
Thanks a lot for any info and advice!!