00:00:12.480
Hello everybody! I'm Paola Moretto, and I'm the co-founder of a company called Nuvola. You can find me on Twitter at @paolamoreto3. A little about me: I'm a developer turned entrepreneur and have been in the high-tech industry for a long time. I love solving hard technical problems. I originally come from Italy, but I've been in the U.S. for 20 years. When I'm not writing code, I'm usually outdoors hiking.
00:00:24.640
Today, I want to talk about performance. We've heard it loud and clear here at RailsConf: faster is better. We all know what performance is, but it's important to understand the real impact of low performance. When I talk about performance, I specifically mean the speed and responsiveness that your application delivers to users. There's a famous quote from Larry Page that says speed is product feature number one. Therefore, you need to focus not only on your functional requirements but also on the non-functional requirements. Speed is paramount for any web application today. There is a lot of research and data backing this up, showing the impact of low performance. For instance, low performance negatively affects your SEO ranking, conversion rates, brand perception, brand loyalty, and advocacy. Additionally, it impacts your costs and resources because the tendency for low performance is usually to over-provision, which isn't the right answer. So, speed is crucial for web applications.
00:02:21.520
If you have a DevOps model, combining development and QA into a single team means that performance becomes even more critical. In the cloud, where you have a fully programmable and elastic infrastructure and are adopting continuous delivery, it’s essential that every build is not only functional but also meets speed requirements.
00:03:10.879
So, what do you do? How do we tackle the performance problem? The first step is to gather data. I came across a quote that I really love: 'In God we trust; all others bring data.' It's a poor model to deploy and then just hope for the best, letting your users become your QA department. This isn't the best approach. We need adequate data to understand our performance issues.
00:03:26.720
There are different types of data to consider. On the right-hand side of your environment, you have your deployments in production and live traffic, which fall under the umbrella of monitoring. Here, you gather various types of monitoring data and techniques. On the left-hand side, you have your testing environment, where you typically have a pre-production or staging environment. Sometimes, testing occurs directly on production. In this setup, you create synthetic traffic for performance testing.
00:05:30.400
Monitoring is the first step of performance management. It involves monitoring your stack, infrastructure, user behavior, and logging. You can also use streaming analytics to get high-frequency metrics. There are many monitoring solutions available, which complement each other depending on your application. However, once you have the data from monitoring, the challenge is to correlate it all and determine exactly what's happening. As the saying goes, you must first instrument and then ask questions. But remember, monitoring alone is not sufficient.
00:06:02.880
Firstly, live traffic is noisy, making troubleshooting difficult, especially if an issue arises during testing. Secondly, monitoring is reactive; it identifies issues after they occur. It's like calling AAA after a car accident; it may help, but ideally, you want to prevent the accident in the first place. Therefore, while monitoring is necessary as your first line of defense, it must be paired with performance testing, which addresses the proactive side of performance management.
00:09:19.039
Gathering synthetic traffic allows you to create controls where you can generate user traffic and conduct performance testing in a pre-production environment to avoid impacting real users. With synthetic testing, you develop scenarios based on user behavior, which provides valuable data for troubleshooting. This method simplifies identifying problematic scenarios, enabling developers to alter traffic and user variables easily.
00:12:28.320
Performance testing generates end-to-end user metrics, providing a clearer picture than server-side metrics alone. A significant disparity can exist between end-user experience and server metrics. So, to understand how performance impacts users, you need the complete view, including user metrics.
00:13:12.320
This data can help diagnose issues early, allowing developers to fix problems before deployment. To achieve this, realistic scenarios must be tested thoroughly. Testing should also encompass different device types. For global applications, performance should be tested from various geographic locations, ensuring comprehensive coverage.
00:14:56.000
When dealing with metrics, the focus should be on measuring response times, transaction completion times, and error rates. The goal is to resolve issues before they reach users and degrade their experience. Understanding when to perform these tests is crucial since software is in constant flux.
00:17:11.680
Applications are also increasingly complex, having multiple optimization points. Frequent testing, especially before implementing changes, prevents blind spots. You should always conduct tests before inversions in traffic to assure expectations align with user experience.
00:19:28.079
However, even with thorough testing, issues may still occur. A common requirement is to measure results across different types of traffic and usage patterns, identifying the first component that slows down. Thorough spending on monitoring alone progresses into identifying solutions through predictive analytics.
00:21:17.759
The next step is data mining and machine learning, which can help localize performance problems. For example, if performance issues arise during testing, analyzing network and external factors such as DNS times can reveal the underlying issues effectively. We want to draw insights from large data sets addressing specific metrics to accelerate troubleshooting.
00:22:57.679
In conclusion, speed is the most critical feature of any application. Performance testing complemented by data monitoring and analytics provides a mathematical approach to identifying performance weaknesses early. Today’s talk demonstrates how combining various methodologies addresses performance concerns before they materialize.
00:23:55.839
Thank you for your time! I would be happy to answer any questions you might have. You can find me on Twitter at @paolamoreto3, and I look forward to your feedback.
