Imagine a large telecom. Many systems, millions of messages carrying sales, service provisioning or payment information, every second, every day. An integration platform of this scale is a complicated system, assembled from parts made by various vendors. Each moment of limited availability means thousands of disappointed clients and directly measurable financial loss. After seven years of operation, an existing solution needed…
a substantial upgrade.
A small agile team had spent almost a year building an integration solution that would suit the needs of its many users for several years to come. The vendors developing the services wanted new technologies, IT operations wanted time-tested ones. Now in a few datacenters over 50 JVM instances run OSGi, Apache Karaf and Fabric8 applications built with CXF, Quartz, Ehcache, Coda Hale Metrics, Oracle UCP, ActiveMQ. They are centrally managed with Zookeeper.
But it's the IT department whose work has changed significantly. Declarative provisioning of JVM containers, configuration managed routing and service versioning simplifies their every day work. Monitoring infrastructre this large requires its own ELK/InfluxDB/Grafana cluster which is a part of solution we've built.