How to write a efficient ES query for the below SQL query?

Elastic Stack Elasticsearch
4

we use es as tsdb, and data we store is look like:

{
	"metric": "metric_name",
	"@timestamp": "2020-01-01T01:01:01",
	"t": {
		"key1": "val1",
		"key2": "val2"
	},
	"f": {
		"name1": 1.2,
		"name2": 2.3
	}
}

simple aggregation is ok (but some aggs are still very slow) like:

select t.key2, sum(f.name1) from tsdata where t.key1='val1' group by t.key2

but i have no idea that if we can handle a more complex aggregation like "join" does in mysql