Some questions regarding metric collection in miq?

I have some questions about metric collection in miq:
(1) when I select utilization in monitor of a vm, the data is displayed hourly at most. This is not very friendly for monitor usage, can this be more precise(eg, by minute)?
(2) the db is postgresql, the performance is very bad when I query the metric(cpu, memory) of vms. Is it ok to store the metrics to time serial db (eg: graphite) for monitor and at the same time to postgresql for the existing functions(eg: report, chargeback).
This make the least changes to miq.
(3) Is there any documents about the deep implement of metric collecting in miq?

@dmetzger can you review this question from @wangxiangyu and forward to a SME if necessary.

(1) @kbrock can you provide input here?
(2) We are working toward using a Time Series Database for use with metrics, I do not have a timeframe for when you can expect this.

We were running into issues around scheduling of the utilization capture. This has been fixed in recent builds. So please update to a recent build.

The realtime captures are rolled up by hour, but you should still have access to the realtime numbers.

configuration value performance.capture_threshold.* determines how frequently we capture vms/hosts/storages. (Typically every 10 minutes, but if the resource has an associated alert, it is capture every 1 minute.)

We are actively looking at converting metrics over to a time series databases. But it is quite a bit of work, so there is no estimate at this time.