PMM performance

Hi Percona engineers,
I’m using PMM to monitor our MySQL servers, I builded a PMM server via docker.Docker and container install on a physical machine with 500G memory, 40 core cpu and SSD disk.

However, when pmm server setup succeed, I added must our DB server on PMM(hundreds of instances). When we access to PMM homepage(<server_ip>:/graph), it will cause very long time to load every instances information, even more than 10 mins and can not show anything. So far we are using a url only including hostname with “pmm” to access and then choose a/some hosts which we want to see.

Have any advices to improve the performance of PMM server? When I create a container for pmm-server, I set the METRICS_MEMORY = 134217728, but the usage of CPU and memory are very low…