Setting Up a Monitoring Strategy in China

In our previous post in this series, we suggested treating China as a second Internet. This concept implies the need to setup a separate monitoring strategy that is tailored to China, rather than just adding it on to your existing strategy. This post provides a step-by-step process to getting a monitoring strategy in China started.

Choose your monitoring locations

When developing a monitoring strategy for the United States, would you choose to monitor from a single city or provider and assume performance was good everywhere based on that limited data point? Probably not. For the same reasons you wouldn’t monitor from a single city in the US, you shouldn’t monitor from a single city or provider in China.

Monitoring should include locations on the three main providers: China Unicom, China Telecom, and China Mobile. Keep in mind that it is possible to be blocked on one carrier and not others. For home broadband, the majority of Internet traffic travels across two of the ISPs, China Telecom and China Unicom. In big cities, China Mobile may be an option for home broadband, but they carry a small majority of traffic compared to China Telecom and China Unicom.

Make sure that mobile monitoring is part of your strategy as well. Some users in China don’t have home broadband and rely solely on their mobile. While China Mobile has a small percentage of home broadband Internet, they are the market leaders when it comes to wireless providers with 63% of the market.

If you’re deploying infrastructure in the cloud with companies like AWS or AliCloud, make sure to choose monitoring locations on those providers’ networks as well.

From a geography perspective, cover the areas of China with the largest consumer bases. A comprehensive monitoring strategy needs to include more than just Shanghai and Beijing, some smaller cities may still have a population of 10 to 20 million people. Ensure you are monitoring from cities both near and far from your hosting point(s).

Run traceroutes and DNS tests

In addition to running standard web or transaction tests, traceroutes and DNS tests can provide valuable information regarding your performance. Knowing whether a timeout, reset, or another error is due to congestion versus the site being blocked can save wasted time trying to resolve the wrong issue. Traceroutes can help shed light on whether issues are due to congestion or another factor.

Traceroutes can also reveal if DNS is hosted in optimal locations. Hosting DNS within mainland China is a good step to avoid potential problems, but it is important to know where the DNS is hosted given the size of mainland China. Some companies have DNS in multiple locations, and optimizing the DNS location may help to improve overall performance.

If DNS issues occur, running detailed DNS tests across top level domain servers, name servers, and root domain servers will help IT teams quickly identify and resolve potential DNS issues.

Chart data by ISP

Given the way in which the two primary ISPs are split geographically, with Unicom serving mostly residents of the north and Telecom serving primarily the south, averaging the data across these is meaningless. The performance between these carriers may vary and if the majority of the users are on Unicom does it matter if the performance on Telecom is better or worse, given they can’t switch. Look at each vendor individually to truly see performance as your users experience it.

Create charts in local China time

Creating charts in the local time zone makes it easier for partners to work with you and help troubleshoot potential issues. When your partners are trying to correlate and understand the data, not having to do additional calculations to convert the time zone makes it possible to resolve problems faster. Making it easier for partners to understand the information will likely motivate them to be more willing to work with you to improve performance and resolve issues when they occur.

Set up different alerts for China

If China is being treated as a second Internet, alerts should be set up independently. Inundation of false positives due to the unique nature of application delivery in China is not a good way to attempt to manage your performance. Consider setting up alerts specific to China-only agents, or alerts on a per-ISP basis. But before setting up any alert, run a baseline to understand the current state of performance from China and don’t assume performance thresholds will be the same as other locations.