微服务调用耗时问题咨询

调用过程nlsb -> edge (服务A) -> provider(服务B)，存在部分（万分之几十） provider 响应只有10ms， 但是edge响应时间800ms左右的请求。edge本身几乎没有处理业务。 类似日志如下
调用时间：2022-12-13 16:24:46.975
一、edge日志：
1. 调用日志
**2022-12-13 16:24:46.975**|transport-vert.x-eventloop-thread-1|DEBUG|LoadBalancerContext.java:getServerFromLoadBalancer:492|c225268d-1e24-4a93-b579-8c15eaa8ae7c|default using LB returned Server: rest://182.10.xx.xxx:8080?sslEnabled=true for request null
2. access.log
2022-12-13 16:24:46.973|182.10.xxx.0=xx|"POST /rtd/kbz_fraud_prevention/kbz_fraud_prevention/P2P_Verify/1.0/verify HTTP/1.1"|200|c225268d-1e24-4a93-b579-8c15eaa8ae7c|**823**

二、provider日志access.log
**2022-12-13 16:24:46.976**|182.10.xx.xxx|"POST /risk-api/v1/kbz_fraud_prevention/P2P_Verify/predict HTTP/1.1"|200|c225268d-1e24-4a93-b579-8c15eaa8ae7c|**10**

edge的CPU及内存使用都不足50%
可能是什么原因导致的呢？

Start by comparing the timestamps and request ID in the edge and provider access logs, then inspect LoadBalancerContext.java around getServerFromLoadBalancer. Trace the interval between the provider response and edge completion using the existing logs. Done means documenting a confirmed cause for the latency gap and the evidence supporting it.

Written by the indexing model from the issue text.