Organizations looking to drive faster time to market, increase agility and deliver a superior user experience for both customers and employees are quickly adopting “Cloud First” strategies. While most enterprises began their cloud journey with a single infrastructure-as-a-service (IaaS) platform, several compelling offerings from the likes of Alibaba Cloud, Amazon Web Services (AWS), Microsoft Azure, Google Cloud Platform (GCP) and others have created a plethora of options, with best-of-breed functionality designed for different workload types. Today, we see many enterprises consuming two or more of these cloud services.
As organizations continue to invest in multi-cloud ecosystems, their reliance on the Internet is growing increasingly mission-critical—yet their control over the underlying infrastructure is shrinking. In a multi-cloud world, the need to monitor these services from a performance and availability perspective is paramount. In response to this growing need from our customers, last year, we announced support for multi-cloud monitoring across the three major public cloud providers: AWS, Azure and GCP. This significantly augmented our vast array of global vantage points, allowing our customers to measure performance to and from these cloud providers, between cloud providers and within (inter-region and inter-AZ) cloud providers.
Introducing Alibaba Cloud Agents for Improved Asia-Pacific Visibility
While enterprises are moving toward multi-cloud ecosystems, those who serve global markets have found that Internet performance in the Asia-Pacific region can fluctuate greatly. In our 2018 Public Cloud Performance Benchmark Report, we noted that in Asia, AWS demonstrated 35% less network performance stability than GCP and 56% less than Azure. And thanks to heavy and opaque sovereign controls over Internet behavior, application performance inconsistencies in China stick out like a sore thumb. The impacts of characteristically unstable Internet performance throughout the Asia-Pacific region is felt particularly by enterprises that serve customers in these markets.
To address this growing concern, we are expanding our fleet of Cloud Agents beyond the 14 agents already deployed in China to include 19 additional Alibaba Cloud regions. This brings ThousandEyes Asia-Pacific vantage points to a total of 53 cities and global vantage points to 184 cities. By adding Alibaba Cloud into this mix, organizations leveraging any combination of Alibaba Cloud, Amazon Web Services (AWS), Microsoft Azure and Google Cloud Platform (GCP) have the ability to measure and visualize app and network-layer performance metrics on a cloud-to-cloud, Internet-to-cloud and inter-region basis.
Understanding Cloud Performance in Asia
To demonstrate the benefits of a multi-cloud approach in certain situations, let’s get a sense of how the four public cloud providers perform by looking at an example of network latency between two major cities in the Asia-Pac region. Figure 1, shown below, represents network latency values across the four public cloud providers for traffic originating in Beijing and terminating in Singapore. In this chart, we can see that Microsoft Azure and Alibaba Cloud out-perform the other cloud providers, and GCP demonstrates a relatively high (compared to Azure and Alibaba Cloud) but consistent latency values. However, what draws the eye in the figure below is the striking lambda wave showing significant fluctuations in latency for AWS. Users of AWS between these locations may be experiencing extremely unstable performance as latency skyrockets and plummets.
To understand what’s happening, let’s compare the path visualizations for both AWS and Alibaba Cloud. In Figure 2 below, which shows traffic across the AWS cloud, you can see traffic is being routed from mainland China to Japan before it is sent to the destination in Singapore. Compared to the network path for Alibaba Cloud shown in Figure 3, this is a circuitous route and probably explains why network latencies were high for AWS to begin with. The occurrence of the lambda pattern (seen in Figure 1) coincided with a new high latency path that was introduced between China and Tokyo. The more interesting observation was that the peaks in network latencies roughly maps to peak time of day in Beijing local time. Network latencies are seen peaking between 8am-8pm Beijing time, and dipping with the lowest latencies around 4-5am Beijing time. This type of time of day peaks and troughs suggests that performance was being affected by traffic congestion during usage rush hour.
Conversely, the path visualization for Alibaba Cloud, seen in Figure 3, shows a much less circuitous route. The paths from the user location in Beijing to Alibaba Cloud's Singapore regions traverses a few nodes in China before reaching the destination in Singapore. This more direct path is likely leading to the faster transmission times and lower latency we're measuring.
Choosing the Best Cloud Approach
While we chose to highlight this as an example of performance differences across public clouds, the main takeaway is to remember that not all cloud providers handle traffic the same way. So when you’re considering a cloud provider or even a multi-cloud strategy, it is important to understand the nuances on a regional level to make the best decision for your business. What works best in one region might not necessarily be best for another, and sometimes the redundancy created by a multi-cloud approach can make up for performance deficiencies in specific markets.
That’s why we’re pleased to expand our multi-cloud monitoring capabilities to include 19 new Alibaba Cloud regions in Asia. Because when you have better visibility, you can make better decisions.
Learn how ThousandEyes can help you monitor your multi-cloud deployments, or request a demo to see it for yourself today. And, if you’re interested in learning more about performance variations across the major public cloud providers, be sure to check out the Public Cloud Performance Benchmark Report.