A faster way than "edge" to retrieve childs

Users Dgraph
1

I’m running this query with 32 cats and 1 million urls takes 3 seconds, that looks excesive to the amount of data

       countsbycat(func: eq(dgraph.type, "**cats**")) @filter(eq(catType, $catType) ) {         
             itemuid: uid
             group: catType
             label: label
            itemsize_eq: count(~**url**.have_tier1 @filter( eq(url.language,"fr")) )

if I change edge reverse and groupby takes similar time

{
       var (func: uid(0x2e8ee66)) {
     	~lang  @groupby(url.have_tier1 ) {    
          a  as  count(uid)      
  			}
      }
   cats(func: uid(a), orderdesc: val(a)) {
    			 itemuid: uid
             group: catType
             label: label
           itemsize: val(a)
  }
      }

infra: 3 zero, 3 alpha in HA/replicated data

Some idea to make this query faster?
Thanks