Elasticsearch VS Easysearch 性能测试

压测环境

虚拟机配置

使用阿里云上规格：ecs.u1-c1m4.4xlarge，PL2: 单盘 IOPS 性能上限 10 万 (适用的云盘容量范围：461GiB - 64TiB)

Easysearch 配置

7 节点集群，版本：1.9.0

Elasticsearch 配置

7 节点集群，版本：7.10.2

监控集群配置

单节点 Easysearch 集群，版本：1.9.0

压测 loadgen 配置

loadgen 版本：1.25.0

4 台压 Easysearch，4 台压 Elasticsearch。

压测索引 Mapping

PUT nginx
{
  "mappings": {
    "properties": {
      "method": {
        "type": "keyword"
      },
      "bandwidth": {
        "type": "integer"
      },
      "service_name": {
        "type": "keyword"
      },
      "ip": {
        "type": "ip"
      },
      "memory_usage": {
        "type": "integer"
      },
      "upstream_time": {
        "type": "float"
      },
      "url": {
        "type": "keyword"
      },
      "response_size": {
        "type": "integer"
      },
      "request_time": {
        "type": "float"
      },
      "request_body_size": {
        "type": "integer"
      },
      "error_code": {
        "type": "keyword"
      },
      "metrics": {
        "properties": {
          "queue_size": {
            "type": "integer"
          },
          "memory_usage": {
            "type": "integer"
          },
          "thread_count": {
            "type": "integer"
          },
          "cpu_usage": {
            "type": "integer"
          },
          "active_connections": {
            "type": "integer"
          }
        }
      },
      "cpu_usage": {
        "type": "integer"
      },
      "user_agent": {
        "type": "keyword"
      },
      "connections": {
        "type": "integer"
      },
      "timestamp": {
        "type": "date",
        "format": "yyyy-MM-dd'T'HH:mm:ss.SSS"
      },
      "status": {
        "type": "integer"
      }
    }
  },
  "settings": {
    "number_of_shards": 7,
    "number_of_replicas": 0,
    "refresh_interval": "30s"
  }
}

压测方法

每 4 个 loadgen 使用批量写入接口 bulk 轮询压测同一集群的 7 个节点，每个请求写入 10000 个文档。

具体请求如下：

requests:
  - request: #prepare some docs
      method: POST
      runtime_variables:
#        batch_no: uuid
      runtime_body_line_variables:
#        routing_no: uuid
#      url: $[[env.ES_ENDPOINT]]/_bulk
      url: $[[ip]]/_bulk
      body_repeat_times: 10000
      basic_auth:
       username: "$[[env.ES_USERNAME]]"
       password: "$[[env.ES_PASSWORD]]"
      body: |
        {"index": {"_index": "nginx", "_type": "_doc", "_id": "$[[uuid]]"}}
        $[[message]]

压测数据样本

{"method":"DELETE","bandwidth":1955,"service_name":"cart-service","ip":"120.204.26.240","memory_usage":1463,"upstream_time":"1.418","url":"/health","response_size":421,"request_time":"0.503","request_body_size":1737,"error_code":"SYSTEM_ERROR","metrics":{"queue_size":769,"memory_usage":1183,"thread_count":65,"cpu_usage":68,"active_connections":837},"cpu_usage":70,"user_agent":"Mozilla/5.0 (iPad; CPU OS 14_6 like Mac OS X) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/14.1.1","connections":54,"timestamp":"2024-11-16T14:25:21.423","status":500}
{"method":"OPTIONS","bandwidth":10761,"service_name":"product-service","ip":"223.99.83.60","memory_usage":567,"upstream_time":"0.907","url":"/static/js/app.js","response_size":679,"request_time":"1.287","request_body_size":1233,"error_code":"NOT_FOUND","metrics":{"queue_size":565,"memory_usage":1440,"thread_count":148,"cpu_usage":39,"active_connections":1591},"cpu_usage":87,"user_agent":"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/14.1.1","connections":354,"timestamp":"2024-11-16T05:37:28.423","status":502}
{"method":"HEAD","bandwidth":10257,"service_name":"recommendation-service","ip":"183.60.242.143","memory_usage":1244,"upstream_time":"0.194","url":"/api/v1/recommendations","response_size":427,"request_time":"1.449","request_body_size":1536,"error_code":"UNAUTHORIZED","metrics":{"queue_size":848,"memory_usage":866,"thread_count":86,"cpu_usage":29,"active_connections":3846},"cpu_usage":71,"user_agent":"Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)","connections":500,"timestamp":"2024-11-16T15:14:30.424","status":403}

压测索引 1 主分片 0 副本

Elastic 吞吐

Elastic 线程及队列

资源消耗

Easysearch 吞吐

Easysearch 线程及队列

资源消耗

对比

压测索引 1 主分片 1 副本

Elastic 吞吐

Elastic 线程及队列

资源消耗

Easysearch 吞吐

Easysearch 线程及队列

资源消耗

对比

压测索引 7 主分片

Elastic 吞吐

Elastic 线程及队列

资源消耗

网络

单节点平均接收 26MB/s，对应带宽：1456 Mb/s

5 千万文档，总存储 105 GB，单节点 15 GB

Easysearch 吞吐

Easysearch 线程及队列

资源消耗

对比

总结

通过对不同场景的压测结果进行对比分析，得出以下结论：

• Easysearch 相比 Elasticsearch 的索引性能显著提升
  Easysearch 集群的吞吐性能提升了 40% - 70%，且随着分片数量的增加，性能提升效果更为显著。
• Easysearch 相比 Elasticsearch 的磁盘压缩效率大幅提高
  Easysearch 集群的磁盘压缩效率提升了 2.5 - 3 倍，并且随着数据量的增加，压缩效果愈发明显。

此测试结果表明，Easysearch 在日志处理场景中具有更高的性能与存储效率优势，尤其适用于大规模分片与海量数据的使用场景。

如有任何问题，请随时联系我，期待与您交流！