Aggregation Sum is very slow

[bruce88 (bruce)]

I have a search on our ES cluster, which has 4 TB, 43 billion docs.

{
"query": {
    "bool": {
        "filter": [
            {
                "term": {
                    "CustomerID": {
                        "value": "s1501"
                    }
                }
            },
            {
                "range": {
                    "Timestamp": {
                        "from": 1533225600,
                        "to": 1535903998
                    }
                }
            }
        ]
    }
},
"aggregations": {
    "UserName": {
        "filter": {
            "bool": {
                "must_not": [
                    {
                        "terms": {
                            "UserName": [
                                ""
                            ]
                        }
                    }
                ]
            }
        },
        "aggregations": {
            "UserName": {
                "terms": {
                    "field": "UserName",
                    "size": 10,
                    "shard_size": 200
                },
                "aggregations": {
                    "total": {
                        "sum": {
                            "field": "InByte"
                        }
                    }
                }
            }
        }
    }
}

}

The search takes more than 10 seconds, but if I remove SUM and only use terms, it only takes about 1 seconds.
Why the SUM takes so much time? Is there any ways to optimize?

BTW, only about 280,000 docs are matched in the search.