Google Compute Engine

Chapter 6. Networking

Fast, responsive networking has been a critical component of Google’s success. Studies have shown that, for a high-volume web business, as little as 100 milliseconds of delay in responding to web requests can have a measurable negative impact on revenue.¹³ Google’s average response time for over 100 billion search queries per month is roughly a quarter of a second, or 250 milliseconds. In addition to the search functionality everyone knows about, Google is constantly moving tremendous quantities of data within and across its worldwide data centers.

Compute Engine is designed to take full advantage of Google’s networking resources and engineering know-how. As a result, whether you’re transferring data between Compute Engine virtual machines, between VMs and other Google Cloud Platform services, or between VMs and the outside world, it’s possible to attain remarkably strong and consistent network performance. In addition to performance, Google Compute Engine offers an impressive array of security, high availability, and configuration options.

In previous chapters, we’ve created virtual machines, persistent disks, and other resources but we haven’t yet described the network environment in which those resources exist. That will be the focus of this chapter.

A Short Networking Primer

Internet Protocol (IP) gurus can safely skip this section. However, those who may be a bit vague on what TCP/IP is, or exactly what a firewall does, may find this section helpful. ...