Performance in the cloud
•Download as PPTX, PDF•
0 likes•302 views
This document discusses the importance of web performance testing and provides guidance on how to test performance in the cloud. It notes that better performance leads to happier users and more revenue. It classifies different types of performance tests and recommends measuring performance before optimizing. The document suggests using both synthetic monitoring tools like WebPageTest to emulate users, and real user monitoring like Application Insights to gather live performance data from actual users. It provides an overview of Application Insights and how it can help optimize application performance on Azure.
Performance in the cloud
- 1. v2016 September 30th 2016v2016 September 30th 2016 Performance in the Cloud NetSuite Sr. Web Performance Engineer Diego Cardozo
- 2. v2016 September 30th 2016v2016 September 30th 2016 Agenda • Why performance matters #perfmatters • What to test • How to test • Performance on the cloud • Test your cloud with local tools • Test your cloud with cloud tools • Let your users do the testing for you
- 3. v2016 September 30th 2016v2016 September 30th 2016 Why performance matters Why is it important for our business? • Google: 2% slower = 2% less searches per user • Yahoo: 400 ms faster = 9% more traffic • Amazon: 100ms faster = 1% more revenue • Google uses site speed in web search ranking Better performance Happier users More $
- 4. v2016 September 30th 2016v2016 September 30th 2016 What to test Classification of performance tests Type of test Goal Performance Determine and validate speed Load Verify application behavior under normal and peak load conditions Stress Verify application behavior beyond normal and peak load conditions Capacity Determine how many users and/or transactions are supported while meeting performance goals
- 5. v2016 September 30th 2016v2016 September 30th 2016 How to test There are only 2 rules 1. Measure first, then optimize 2. Repeat rule number 1 frequently
- 6. v2016 September 30th 2016v2016 September 30th 2016 What to test Have a goal – 3 alternatives • Performance Budget • Twitter uses “time to first Tweet” • Google created the RAIL model
- 7. v2016 September 30th 2016v2016 September 30th 2016 How to test 2 approaches • Synthetic monitoring Emulate a user's browser, run a test case and register response times for all requests • Real User Monitoring (RUM) Users passively send real performance data
- 8. v2016 September 30th 2016v2016 September 30th 2016 </talk> time to see some real tools…
- 9. v2016 September 30th 2016v2016 September 30th 2016 Synthetic monitoring • Quick performance tests WebPage Test (free) • Add load tests to your local dev suite Visual Studio Enterprise Load tests • Run load tests from the cloud (free/paid) Visual Studio Team Services
- 10. v2016 September 30th 2016v2016 September 30th 2016 Real User Monitoring (RUM) Application Insights – available for free on Azure • Gather performance data from real users • 360° view of your application • Server level – IIS • Application level – SDK Telemetry data • Client level – data collection through JavaScript
- 11. v2016 September 30th 2016v2016 September 30th 2016 Application Insights • Application performance monitoring Measure performance, discover failures • Interactive data analytics Ad-hoc queries and deep diagnostics • Proactive detection Machine learning based anomaly detection • Not just .NET Java, Ruby, Python, PHP and Node.JS on Azure
- 12. v2016 September 30th 2016v2016 September 30th 2016 Resources and questions • ݺߣs http://bit.ly/netconf-performance • RAIL http://bit.ly/rail-performance • Microsoft’s performance testing guide http://bit.ly/microsoft-performance dcardozo@netsuite.com @diecard