Grafana and Openstack : Plugin/database architecture choice

Thinking of using Grafana to improve monitoring of an up and running Openstack Queens installation - we would want to connect it to various existing databases on the controllers : Gnocchi (VM metrics) , Zabbix (server PM metrics) and ELK (logs)

  1. In general do we need to install any agents/plugins on the Openstack controllers , or do we just need to add the available Grafana plugins to poll each database ?
    I assume the Grafana plugins are read-only towards these 3 d.b sources?
    Any modification of openstack configs files needed (CORS ?)

  2. I’ve nearly always seen a combination of Grafana + a TSDB (e.g Prometheus/Influxdb) - is it essential to install a TSDB here (for performance reasons)?
    My fear is, if no TSDB installed (relying on just the Grafana plugins) , I try to view a week’s data and get a long wait for Grafana to poll all that data from the controller databases and refresh the page . Are my assumptions correct here ?

  3. Any links towards how to dimension the Grafana installation in terms of metrics/sec , storage requirements ?

thanks in advance for your time and feedback,