---

概述

继续跟中华石杉老师学习ES，第48篇

课程地址： https://www.roncoo.com/view/55

---

官方说明

Percentiles Aggregation： 戳这里

---

示例

需求： 网站访问时延统计

为了演示该功能，我们先建个索引，模拟一批数据

#新建索引
PUT /website
{"mappings": {"logs": {"properties": {"latency": {"type": "long"},"province": {"type": "keyword"},"timestamp": {"type": "date"}}}}
}

#模拟数据
POST /website/logs/_bulk
{ "index": {}}
{ "latency" : 105, "province" : "江苏", "timestamp" : "2018-10-28" }
{ "index": {}}
{ "latency" : 83, "province" : "江苏", "timestamp" : "2018-10-29" }
{ "index": {}}
{ "latency" : 92, "province" : "江苏", "timestamp" : "2018-10-29" }
{ "index": {}}
{ "latency" : 112, "province" : "江苏", "timestamp" : "2018-10-28" }
{ "index": {}}
{ "latency" : 68, "province" : "江苏", "timestamp" : "2018-10-28" }
{ "index": {}}
{ "latency" : 76, "province" : "江苏", "timestamp" : "2018-10-29" }
{ "index": {}}
{ "latency" : 101, "province" : "新疆", "timestamp" : "2018-10-28" }
{ "index": {}}
{ "latency" : 275, "province" : "新疆", "timestamp" : "2018-10-29" }
{ "index": {}}
{ "latency" : 166, "province" : "新疆", "timestamp" : "2018-10-29" }
{ "index": {}}
{ "latency" : 654, "province" : "新疆", "timestamp" : "2018-10-28" }
{ "index": {}}
{ "latency" : 389, "province" : "新疆", "timestamp" : "2018-10-28" }
{ "index": {}}
{ "latency" : 302, "province" : "新疆", "timestamp" : "2018-10-29" }

DSL:

#网站访问时延统计（50% 95% 99%的请求）
GET /website/logs/_search
{"size": 0,"aggs": {"latency_percentiles": {"percentiles": {"field": "latency","percents": [50,95,99]}},"latency_avg": {"avg": {"field": "latency"}}}
}

返回：

{"took": 3,"timed_out": false,"_shards": {"total": 5,"successful": 5,"skipped": 0,"failed": 0},"hits": {"total": 12,"max_score": 0,"hits": []},"aggregations": {"latency_avg": {"value": 201.91666666666666},"latency_percentiles": {"values": {"50.0": 108.5,"95.0": 627.4999999999997,"99.0": 654}}}
}

可知,

• 50%的请求 在 108.5毫秒
• 95%的请求 在 627.5毫秒
• 99%的请求 在 654毫秒
• 平均时延 201.9毫秒

数值的最大的值是多少，不是完全准确的

---

需求二：按地区分组后看下每个地区的 网站访问时延统计

GET /website/_search
{"size": 0,"aggs": {"group_by_province": {"terms": {"field": "province"},"aggs": {"latency_percentiles": {"percentiles": {"field": "latency","percents": [50,95,99]}},"latency_avg": {"avg": {"field": "latency"}}}}}
}

返回：

---

Percentiles优化 compression

compression 官方指导： 戳这里

TDigest算法，用很多节点来执行百分比的计算，近似估计，有误差，节点越多，越精准

compression ，限制节点数量最多 compression * 20 = 2000个node去计算 ，默认100

compression 越大，占用内存越多，越精准，性能越差

一个节点占用32字节，100 * 20 * 32 = 64KB

简单来说： 如果你想要percentile算法越精准，compression可以设置的越大