从 Azure blob 容器接收分块数据时数据包之间的延迟

如何解决从 Azure blob 容器接收分块数据时数据包之间的延迟

我们正在开发一种物联网解决方案，该解决方案将从 Azure blob 容器下载固件和配置文件。嵌入式设备使用 Winc3400 来运行 Wifi 堆栈。 Winc 堆栈对每个套接字发送/接收操作的限制为 1440 字节。

我们看到的是，每 10-12 个数据包，我们就会收到一个短数据包，然后延迟 5 秒。这导致大文件的下载时间很长。

当我们从 AWS 下载任意文件时，我们看不到这一点。

这是下载过程中的日志片段：

[02/04/21 - 12:36:24:185] entity: received received size 930  
[02/04/21 - 12:36:24:213] entity: received received size 1446  
[02/04/21 - 12:36:24:241] entity: received received size 1446  
[02/04/21 - 12:36:24:270] entity: received received size 1446  
[02/04/21 - 12:36:24:296] entity: received received size 1446  
[02/04/21 - 12:36:24:367] entity: received received size 1446  
[02/04/21 - 12:36:24:418] entity: received received size 1446  
[02/04/21 - 12:36:24:446] entity: received received size 1446  
[02/04/21 - 12:36:24:474] entity: received received size 1446  
[02/04/21 - 12:36:24:502] entity: received received size 1446  
[02/04/21 - 12:36:24:534] entity: received received size 1446  
[02/04/21 - 12:36:24:560] entity: received received size 491  
[02/04/21 - 12:36:29:915] entity: received received size 1433 <- 5 seconds since last packet

我一直无法弄清楚为什么我们会在数据包之间出现这些延迟。堆栈在 MCU 外部的 IC 上处理，因此我们深入研究它的能力有限。

Azure 和 AWS 容器之间是否存在任何可能导致这种情况的主要差异？

这可能只是 Azure 服务器和我们设备上有限的套接字传输大小的问题吗？

我们可以在 GET 请求标头中放入任何有助于解决问题的内容吗？现在我们只指定“传输编码：分块\r\n”和“连接：保持活动\r\n”。

谢谢！