database partitioning strategy for time-series data

Hey everyone, So, we've got a ton of time-series data piling up for our monitoring metrics and logs. Our current setup, with just one giant table, is starting to really slow down. We're thinking about partitioning, but I'm not sure what the best way to go about it is. What partitioning keys have you found work well for time-series stuff? Like, should we go by day, week, month, or maybe a rolling window? Also, how do you manage keeping old data around when you're using partitions, and does it mess with query speeds if you need to look across a bunch of them? Really interested to hear what you all have tried and what you suggest!