提升网络规格以在 GCP 计算引擎实例上更快地联网?

Upsize network specs for faster networking on GCP compute engine instances?

我有一个计算引擎实例 运行,我 运行 一些 bigquery 查询,现在将结果集下载到计算引擎实例以进行进一步的工作和探索。预计到达时间为 5 小时,实例每小时花费几美元,因此为更好的网络功能支付更多费用并尽可能减少 t运行sfer 时间符合我的利益。

问题

是否有可能 select 计算引擎实例上更高的网络规范,如果可以,怎么做? (即 cli 命令/选项是什么,或者,如何通过 GCP 控制台实现)

遗憾的是,无法通过 Cloud Console 或 SDK 在 GCE VM 上 select 更高的网络规格,但您可以在文档中找到一些有用的信息。

先看文档Machine types section Network bandwidth:

Google Cloud accounts for bandwidth at the VM level, not per network interface (NIC) or IP address. Neither adding more IP addresses to a NIC with alias IP ranges or creating a VM with multiple NICs will increase either inbound or outbound bandwidth.

Inbound bandwidth Google Cloud handles inbound traffic to a VM differently depending on whether the packet's destination is an internal IP address or an external IP address.

  • To an internal IP address Google Cloud places no restriction on traffic inbound to an associated internal IP address. A VM is allowed to receive as much internal traffic as its machine type, operating system, and other network conditions and resources permit.

  • To an external IP address Google Cloud limits inbound traffic sent to a VM's associated external IP address to whichever of the following rates that would be exceeded first:

    • 1,800,000 packets per second
    • 20Gb/s

Note: Depending on the machine type of your VM and other factors, actual inbound traffic might be less than 20Gb/s. Bandwidth from the internet, for example, is not covered by any SLA and is subject to network conditions.

另一方面,outbound bandwidth 取决于 VM 的机器类型。

之后,看看这篇文章5 steps to better GCP network performance

此外,我建议您 select 适当地 storage 让您的 GCE VM 达到预期的网络带宽:

Block storage resources have different performance characteristics. Consider your storage size and performance requirements to help you determine the correct block storage type for your instances.