From our partners at Neotys
You always want to know that your website is operating at its best, but how do you know that’s actually the case? It’s not so easy to see behind the curtain when it comes to your web infrastructure. We’ve long used proxy metrics like CPU load or server availability to ensure that a server is “up,” but these measurements don’t provide enough data. In fact, as websites become more complex and change more frequently, these measurements become less useful.
A website visit may involve a wide range of components many of which are off-site or not easily monitored. External ad servers, web service APIs, content delivery networks, and even specialized back-end systems – each of these represent a potential bottleneck that could impact an important transaction without raising an appropriate red flag.
So what do you have in your toolbox to help address this? Load testing, real-user monitoring and site instrumentation all help you prepare for and monitor your website visitors’ experiences. But one more tool that’s essential for a performance engineer is synthetic user monitoring. It’s a critical part of a web monitoring strategy, however for many people, it’s uncharted territory. So, in this post we want to show you what is required for proper synthetic user monitoring.
Simply put, synthetic user monitoring lets you create virtual users that operate externally to your system and mimic real user behavior by running through user paths on your website or application, while you measure the performance of the system. You do this while the application is in use, on a live production system. Why? Because that’s how you can see what your users are seeing, without requiring real users to execute those tasks.
Take this example: you have a check-out cart on your site – a high-value transaction, and therefore one that deserves a flawless experience. Not everyone gets to the cart. Most people are browsing the rest of the site. But when people do get there, you want to make sure they have an amazing experience.
If you rely only on measuring that experience only when a user is actually checking out, you have no way of knowing what that experience will be like for that user. You put the high-value transaction in jeopardy because you don’t have any data about how well it will perform until there is a real person going through it.
This is exactly what synthetic users are for. You build a simulated transaction that mimics a user’s most common tasks: add to cart, checkout, log in, etc. As load increases on the production site and more and more visitors get ready to buy, you can continually check to see what the experience is like along those key tasks without putting any individual visitors at risk. That way, you know about problems your users might encounter before they do.
So what should you be looking for in a Synthetic User Monitoring tool? Here are six attributes that should definitely be on your list of key requirements for synthetic user monitoring.
Synthetic users are great for application transactions that really matter, and these user paths are rarely simple. Your synthetic user monitoring solution should include support for interacting with and navigating through a wide range of web technologies: Flash, HTML5, Google SPDY, Push, WebSocket, and any other of the latest web and mobile technologies.
With this support under your belt, you aren’t limited in how you put synthetic users to work. Take a look at your web analytics and find out what your most common paths through the site are. Then recreate those in your synthetic user monitoring tool, exactly as your users experience them.
Beyond that, think about how you can leverage synthetic testing for new features, before you let real users in. Deploy a feature on a special build that’s running on the same server. Don’t direct live users to it yet, but plan some paths for synthetic users. Then at times of peak load, run the synthetic users through the script to see how the experience is.
There are plenty of other ways of leveraging synthetic user monitoring to be more proactive. By thinking about the future first, you’ll use synthetic user monitoring to its maximum benefit. Check out more tips here.
Once it’s set up, synthetic user monitoring is a fantastic tool. What holds many people back is being able to write a script that lays out an entire decision and process tree that a user could make. So you want a tool that makes this as easy and frictionless as possible.
As stated above, you want to create scripts that are modeled after real user behavior. A no-coding solution for script development makes this process significantly easier because you work within a graphical interface, putting blocks of functionality together without the pitfalls and complexities of manually written scripts.
You can also incorporate other attributes of user behavior into your scripts – for example, connection speeds and browser behaviors. You can execute scenarios from various geographies for further realism in your testing.
A no-coding solution for test scripts means you can quickly churn out a robust, representative library of tests that will accurately simulate your users.
You gain a lot of efficiencies by reusing your load testing cases in your synthetic user monitoring tool. If you think about, there isn’t much difference between what you want to test in a load test and what you want to test in synthetic user monitoring. In both cases, you are looking to leverage realistic test scenarios to see how the system behaves before a real user experiences a problem.
So repackage your load tests as synthetic user monitoring tests, and look for a tool that allows you to share them between these different testing environments. You’ll want to be able to easily port your load tests into synthetic user monitoring tests, and you may even find that a new synthetic user monitoring scenario would make a really good structured load test. You can often use the same data too, which is a great way to test in production and test in the cloud without putting data at risk.
Your mom always told you to recycle. Here’s just another way to do that!
A good synthetic user test will simulate a real user as accurately as possible, and one key characteristic of that experience is the network. Not everyone connects to the Internet with the same high-quality connection. You’ll want a synthetic user monitoring tool that emulates various network speeds (3G, 4G, Wi-Fi) as well as network errors like packet loss and latency.
When everything works smoothly, users are likely to have a good experience. But things don’t always go smoothly – that’s when errors occur and users complain. How does your application perform in the face of these errors? That’s a key question you’ll want to ask and one of the ways you can leverage a modern synthetic user monitoring tool.
Introduce errors into your test scenarios to see how your app behaves under stress. If there is a network error along the way, do client apps suddenly start drawing down lots of data as part of a re-sync protocol? What happens when this takes place at scale? The data you collect through your synthetic user monitoring tool has a tremendous amount of value and can help improve the system – and the user experience – in many ways.
If you haven’t gotten the memo yet, web users are mobile. You should no longer be thinking about these as two separate environments or even two separate user bases. Today, the rule is “mobile first.” So you need to be monitoring both your mobile users and your desktop users as a common set of visitors.
Your synthetic user monitoring tool should have the ability to emulate a wide range of mobile devices so you can determine how those users may be experiencing your website, and particularly if there are any differences between what someone sees on their phone as compared to their computer.
Be sure to consider mobile load testing and monitoring right from the start, when setting up your initial synthetic monitoring and tests. Leverage your analytics data to find out how many users are on mobile and what they are doing. Don’t treat this as secondary – today’s web users are on their devices, maybe even more so than their computers.
Are you still trying to figure out if a dedicated mobile testing environment is important? Check out our infographic, Mobile-First Performance – it may persuade you.
Make sure your synthetic user monitoring solution takes advantage of all the information available and makes it accessible through a rich set of dashboards and real-time notifications. You should have access to real-time and historical data, along with the ability to set and monitor key performance indicators (KPIs). You’ll also want to configure alerts so your monitoring team can take action when SLAs are violated.
This is a critical requirement, as it turns your synthetic user monitoring tool from a learning system to a doing system. Regular synthetic tests can monitor performance and immediately alert staff to fix a problem before a user experiences it.