Graphene-python大数据集处理性能问题求助

我最近在项目里用 graphene-python 搭配 graphene-django（还加了 graphene-django-optimizer 做数据库查询优化），碰到了个非常棘手的问题：

• 数据库查询部分表现完全正常，所有SQL加起来也就数毫秒就能跑完；
• 但graphene把查询结果转换成GraphQL响应却要花10秒以上；
• 而且响应数据量和延迟呈线性正相关——数据量翻三倍，响应时间也跟着翻三倍。

返回的数据是最多7层的嵌套对象，数据库查询已经通过optimizer优化过了，SQL耗时完全没问题，所以我百分百肯定延迟出在graphene的结果解析/序列化环节。

我试过用Django的cProfiler来抓性能，但不知道为什么没法定位到graphene的执行过程；后来用 graphene-django-debugger 中间件拿到了SQL查询的耗时数据，大家可以看看：

"_debug": { "sql": [ { "duration": 0.0016078948974609375, "isSlow": false, "rawSql": "SELECT SYSDATETIME()" }, { "duration": 0.0014908313751220703, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.0014371871948242188, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.001291036605834961, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.0013201236724853516, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.0015559196472167969, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.0016672611236572266, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.0014820098876953125, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.0014810562133789062, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.001667022705078125, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.0014202594757080078, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.0027959346771240234, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.002704143524169922, "isSlow": false, "rawSql": "SELECT [redacted]" }, { "duration": 0.0030939579010009766, "isSlow": false, "rawSql": "SELECT [redacted]" } ] }

同一请求的服务器响应时间有截图佐证（截图位置：[此处]）。

有没有大佬遇到过类似的情况？或者能教教我怎么才能对graphene的序列化过程做精准的性能分析？实在是卡在这里动不了了，万分感谢！

内容的提问来源于stack exchange，提问作者techknowfile