Cloud Networking Matters

ThousandEyes recently published the Cloud Report 2022. A few days before that I was fortunate enough to get a pre-briefing on it from Angelique Medina, Head of Internet Intelligence and Director of Product Marketing at ThousandEyes. That gave me a few days to consider the implications of what she told me. Much has changed since the last ThousandEyes Cloud Report published in 2019. For a start ThousandEyes was acquired by Cisco in 2020 For businesses, the global Covid19 pandemic accelerated an already growing dependency on digital services, driving increased cloud-first strategies and a continued expansion of public cloud networks. We had already identified the potential visibility gap in monitoring end-to-end network performance in our 2019 White Paper Monitoring and managing the performance of complex hybrid IT infrastructure environments, but this report from ThousandEyes provides empirical evidence that exposes some of the potential risks businesses need to be aware of when much of their revenues and profits rely on public networks they don’t own and, into which, they often have little visibility.

The report delves extensively and deeply into three core sets of network performance data; end-user measurements, inter-region measurements and inter-availability zone (AZ) measurements for Amazon Web Services (AWS), Microsoft Azure, and Google Compute Platform (GCP). The breadth and depth of ThousandEyes monitoring of public networks is quite remarkable and has resulted in a report full of fascinating and critical insights that all organisations need to study and understand. I’ll leave you to pour over the very detailed report, but there are a number of key findings I think are important to highlight here.

Cloud Providers have their own business imperatives

That is hardly a surprise. However, those imperatives may lead them to take decisions about routing, peering, and rolling out new features only to certain regions first (as an example), or even certain types of traffic. These could have significant performance implications for end users.

There is no “steady-state” in the Cloud

This ought to be obvious. The huge growth in both the volume and variety of cloud traffic forces the cloud providers to constantly evaluate how they build out and operate their global networks. This can lead to frequent and significant changes in latency and overall performance in different parts of the world at different times. Taking point-in-time snapshots of your cloud networks is no longer sufficient.

Cloud network performance issues are common

Mercifully, large scale outages that make global headlines are relatively rare. On the other hand, the cloud providers, like any other business with a private network, will experience fairly regular performance issues and even local outages. Our own anecdotal evidence suggests that the lack of major outages has made many businesses complacent about the implications for them of such performance issues. You should have a clear understanding about how you can mitigate, for example, a complete AZ outage, or how you could re-route traffic if your cloud provider’s routing changes cause unacceptable latency and packet loss issues.

In summary, Having this information should assist businesses in architecting their applications appropriately for the environment they will be running in.

This report serves to emphasise that, by and large, you can, and should, monitor cloud and other public networks. I also think it shows that how you architect your multi-cloud and hybrid-cloud networks, is now a critical business imperative.