Karavel.com is the number one travel site in France, receiving 4 million unique visitors a month and helping millions of travelers reach their destinations around the world.
Over a decade old, the company had to implement major changes in order to achieve scale and give its end users a seamless experience. It transitioned from a monolithic WebLogic architecture to a SOA/ distributed environment that included over 200 Tomcat servers, multiple databases, and asynchronous web services communicating to multiple web partners. The site processes about 3,000 transactions per minute.
But such a complex architecture meant that the Operations and Development teams at Karavel found themselves facing a variety of issues, including site performance problems as a result of non-optimized SQL queries, open source framework performance bugs, and resource leakages. They attempted to resolve these issues using logs as well as thread and memory dumps, but ended up with an average Mean-Time-to-Resolution of over three hours.
Guillaume Postaire, Infrastructure Architect for Karavel noted, “Even after we resolved whatever problem we were dealing with, we still weren’t finished firefighting. The resulting post-mortem analysis could take as long as a day. That’s why we decided to go out to the market and evaluate application performance management solutions.”
Compuware dynaTrace and CA Wily Evaluated But Deemed "Hard to Use" and "Costly"
Karavel’s team looked at the CA Wily Introscope product, but dismissed it as being too difficult to configure and use. Dynatrace came next—but Dynatrace refused to give Karavel employees logins to the product to configure and install it themselves during the Proof Of Concept. Only Dynatrace consultants were allowed to configure and use the software. Karavel employees were only allowed to watch the results of the Dynatrace engineers, which led Karavel to conclude that the product was too difficult to use.
In addition, Dynatrace refused Karavel’s request to test the tool in Karavel’s production environment; they insisted that the evaluation be conducted in a QA/Test setting. This also led Karavel to conclude that the product was not designed for Production environments.
Benoit Villaumie, Karavel’s Lead Architect, remembered: “We couldn’t make a buying decision without having a Proof of Concept in our production environment, which is where we were worried about visibility and root cause analysis. When Dynatrace refused to allow this, we decided to search the web and find their competitors. We found AppDynamics and asked for the 30-day trial license—which we received immediately after doing an online demo with a sales representative. AppDynamics had no problem giving us the trial license keys and logins and letting us do our own deployment in our production environment.”
Guillaume Postaire added, “From the first moment with AppDynamics, it was clear that they had designed an intuitive product that was intended for customers to self-serve and enjoy using to monitor production environments. And we never looked back. AppDynamics delivers great results while being extremely easy to use.”
The AppDynamics Experience
Karavel deployed AppDynamics in production without having to rely upon any consultants. Upon doing so, the Karavel team immediately identified a production performance bottleneck due to bugs in the Java framework, thanks to AppDynamics’ ability to auto-discover and display full call stack latency breakdown for problematic business transactions.
Karavel also identified a performance regression problem after a production release was deployed, due to AppDynamics’ ability to automatically baseline the performance of business transactions and compare performance across releases.
Even system analysis was vastly simplified since the team could now set up custom dashboards and review metrics related to memory and garbage collector activity. Root cause analysis became much faster and much less painful compared to the previous practice of analyzing log files and thread dumps.
Benefits with AppDynamics
Mr. Postaire said, “AppDynamics is now used by around 50 people that include business analysts, developers, architects and operations, but it’s so easy to use anyone can log in and figure it out—no one needs to ‘own’ it. Business, Ops, Dev, and the Architecture team simply use the solution as needed.”
Crucial business transactions, such as the “search” transaction—where end-users review potential transportation options—have improved dramatically. “We’ve seen a 20% increase in performance across many of our key transactions,” said Mr. Postaire.
Mr. Villaumie pointed out that AppDynamics has also greatly reduced Mean-Time-to-Resolution when problems occur. “We no longer spend hours searching for root cause. Typically now it’s only minutes.”
The bottom line: AppDynamics provided tremendous value while being extremely easy to use. “AppDynamics was approachable and made the entire process easy,” Mr. Postaire said. “And the value they’ve delivered to Karavel is tremendous. Our entire team loves using their solution.”
|Example Use Case||Before AppDynamics||After AppDynamics||Benefits|
|Reduction in Severity 1 incidents from optimizing application||30 incidents a year||11 incidents a year||Reduced Severity 1 incidents by 63%|
|Reduction in production downtime||99.90% availability||99.97% availability||Reduced downtime by 72%|
|Reduction in average MTTR per Severity 1 incident||15 hours per incident||7 hours per incident||Reduced troubleshooting time by 50%|
|Total Savings from increased uptime and availability in first year:||€294,306 ($379,566)|
|Projected ROI (4x) over 2 years:||€588,612 ($759,132)|
|Improved performance of key transactions||Several slow transactions impacted end user experience||20% increase in performance of key transactions||Enhanced end user experience and service|
|Sharing responsibility for application performance||Only a few individuals tackling performance incidents||Over 50 users of AppDynamics||Better collaboration and alignment between development, operations and the business|