why my elasticsearch search slow? - elasticsearch

Question

I have the following mapping in elasticsearch5.5:
"append": {
"type": "short"
},
"comment": {
"type": "text",
"index": "analyzed"
},
"create_time": {
"type": "date"
},
"desc_score": {
"type": "integer"
},
"goods_id": {
"type": "long"
},
"goods_name": {
"type": "keyword",
"index": "not_analyzed"
},
"handler": {
"type": "keyword",
"index": "not_analyzed"
},
"is_first_review": {
"type": "short"
},
"logistic_score": {
"type": "integer"
},
"mall_id": {
"type": "long"
},
"order_id": {
"type": "long"
},
"order_sn": {
"type": "keyword",
"index": "not_analyzed"
},
"parent_id": {
"type": "keyword",
"index": "not_analyzed"
},
"review_id": {
"type": "long"
},
"score": {
"type": "integer"
},
"service_score": {
"type": "integer"
},
"shipping_id": {
"type": "long"
},
"shipping_name": {
"type": "keyword",
"index": "not_analyzed"
},
"status": {
"type": "short"
},
"tracking_number": {
"type": "keyword",
"index": "not_analyzed"
},
"update_time": {
"type": "date"
},
"user_id": {
"type": "long"
}
When I profile in kibana with the following statement and got the result in the picture.
{
"from": 0,
"size": 10,
"query": {
"bool": {
"filter": [
{
"terms": {
"goods_id": [
"262628158"
],
"boost": 1.0
}
},
{
"terms": {
"status": [
"2",
"4"
],
"boost": 1.0
}
},
{
"range": {
"create_time": {
"from": "1514027649",
"to": "1514632449",
"include_lower": true,
"include_upper": true,
"boost": 1.0
}
}
}
],
"disable_coord": false,
"adjust_pure_negative": true,
"boost": 1.0
}
},
"sort": [
{
"create_time": {
"order": "desc"
}
}
]
}
I am confused why status filter cost so much time, and status is a field with only five value, this is perpahps the reason cause the problem but I do not know how to optimize my search clause. I have search a lot with google but got no answer yet. Any one can help me?

Answer 1

Finally I find the answer, it is because elasticsearch5.5 use kd-tree to store index data for numeric data type(short, long, etc.) for optimize range search which can be found in following article:
Better Query Planning For Range Queries
Tune For Search Speed
Elasticsearch Query Execution Order
Kd-tree is not suitable for exact term, and I changed the filed type to keyword.