Application Performance Monitoring at Scale

Why implement Lead Wire APM ?

Lead Wire is a DevOps tool for IT operations teams and developers. its implementation makes it possible to achieve several goals :

• Creating internal efficiencies and synergies between operational teams: Lead Wire offers reliable metrics, real-time alerts and granularity of fine analysis; enabling teams to collaborate effectively to resolve incidents.
• Improve service quality and customer relations: Real-time monitoring of the user experience makes it possible to quickly realize the slightest degradation in the service; Problems are detected and corrected before support teams are solicited by users.
• Improve business results (conversion rate): Critical business transactions are monitored; it is the best guarantee of the proper functioning of your company’s  key activities.

Lead Wire provides a complete response to application performance issues; it combines heterogeneous and complementary monitoring techniques :

• Real User Monitoring : Measures user experience (javascript agent injection in web pages)
• Application Monitoring : Applications are instrumented (by agents) to add traces of execution of operations.
• Network Monitoring : Network exchanges are analyzed to track transactions
  
• Synthetic Monitoring : Active monitoring to ensure the health of the application at all times (even in the absence of actual users)
• Infrastructure Monitoring : Collection of system and network resource consumption measures (Lead Wire does not replace the infrastructure supervision tools but covers at least the technical aspects underlying the execution of the monitored applications)

Innovative architecture at the service of performance.

Lead Wire uses newest technologies to make the most of the data collected:

• Elastic architecture : the horizontal scalability allows a very large increase in load. Consolidation of the large mass of data is ensured by the ELK stack.
  
• Architecture Discovery : a system of distributed traces make it possible to reconstitute in a graphic way the application components and their relations
• Anomalies Detection : algorithms (machine learning) analyze in real time the measurements sent by the agents (time series) in search of anomalies or change of trend

We adopted a very strict test protocol to validate our Java agents. The main idea is to control the overhead generated by the instrumentation:

• The microbenchmark : MooBench (MOnitoring Overhead BENCHmark) measures and quantifies the overhead of the monitoring frameworks.
• The macro-benchmark : The SPECjEnterprise® 2018 Web Profile benchmark measures full-system performance of Java EE Web Profile application servers and their supporting infrastructure such as JVM, Database, CPU, disk and servers.