Cloud Performance Varies Among Providers, Across Regions

There can be a good deal of variation in performance across cloud providers, the 2019-2020 edition of ThousandEyes' Cloud Performance Benchmark report found.

The 72-page report is based on a study that looked at Amazon Web Services (AWS), Google Cloud Platform (GCP), Microsoft Azure, Alibaba Cloud and IBM Cloud over a period of 30 days from a number of vantage points. ThousandEyes collected 320 million data points from 98 global metro locations to gauge performance across the public cloud services.

The report found that both AWS and Alibaba Cloud rely on the public internet for much of the data transport, which means there is the potential for unknown risks to cloud performance to be introduced as neither company owns the data links. AWS does have its own private network, known as the AWS Global Accelerator, though apparently it doesn't always outperform the public internet.

"When AWS launched its Global Accelerator in November 2018, the intent was to let customers use the AWS private backbone network for a fee—rather than use the public internet, which is AWS’ default behavior," Angelique Medina, ThousandEyes' director of product marketing, told IT Pro Today. "While there are many examples in various regions around the world where the Global Accelerator trumps the internet connectivity path in performance, the ThousandEyes Cloud Performance Benchmark found examples of negligible improvement—and even cases of worse performance—when compared to default AWS connectivity via the internet."

Typically the cloud providers either use public internet links or private links, with IBM being the only cloud provider with a hybrid approach to cloud connectivity from users to hosting regions. According to ThousandEyes, depending on the hosting region and the expanse of the IBM Cloud backbone, user traffic rides the internet longer or enters the cloud provider’s backbone closer to the end user.

Geographic Latency

Looking at global cloud performance, ThousandEyes found roughly similar levels of performance across Western Europe and North America. The same cannot be said for other regions, with GCP having 2.5 to 3 times the network latency of its rivals, when measuring connectivity from Europe to India.

There is also a dramatic reduction in performance for all cloud traffic heading into China, thanks to China's so-called "Great Firewall," which is a robust content filtering machine.

"Employing a multitude of censorship tools—such as IP blocking, DNS tampering and hijacking, deep packet inspection, and keyword filtering—the Great Firewall is designed to ensure that online content aligns with the government party line," the report states. "Privacy and ethics concerns aside, one of the drawbacks to this system is a vast reduction in performance."

Internet Insights

ThousandEyes is in the business of providing network-level visibility for its users. Alongside the cloud benchmark report, the company announced its new Internet Insights service, which goes beyond what it had been offering in the past.

Medina explained that ThousandEyes' core service provides cross-layer visibility into application delivery over the internet. In contrast, she noted that Internet Insights is service provider-centric and provides a broad view of internet health. It leverages telemetry data derived from the testing performed by all of ThousandEyes' customers to identify outage events in service provider networks.

"Internet Insights is highly complementary to our existing offering, as it enables our customers to understand their application delivery in the context of the wider internet," Medina said. "It also enables our customers to manage external providers more effectively because they now have historical visibility into availability issues—not just globally, but regionally as well."

The improved visibility has already been a big help to one ThousandEyes customer. The customer was complaining that it was unable to connect to a service, according to Medina. It could see in its ThousandEyes tests that there was network packet loss in an upstream telecom provider but couldn’t determine the scope of the issue and why so many customers appeared to be impacted.

"Using Internet Insights, they were able to trace the cause to widespread internet issues that were caused by a Cloudflare route leak," she said. "They were able to mitigate the impact of the route leak early enough to get ahead of the issue by communicating with customers and working with one of their providers to reroute traffic around the most significantly impacted zones."