In the early days, it seemed like a great idea to deploy one application on a Windows server. But then data centers were littered with servers and the practice became unwieldy. Virtualization solved the problem. A great many virtual machines (VMs) could be packed onto one physical machine. Innovation in virtual servers led to a massive wave of hardware consolidation as well as a greater simplification of management. But as the number of VMs mushroomed, VM sprawl became the next barrier. VMs were so easy to spin up that nobody knew how many there were, where they were located, and whether they even needed to be running at all. Gradually monitoring and management tools evolved to help IT.
Now we have containers moving to the forefront and even threatening to unseat VMs as the best way to deploy applications and workloads in a cloud-native infrastructure.
According to Gartner, more than 85% of global enterprises will be running containerized applications in production by 2025, up from less than 35% in 2019. But as the number of containers has exploded, problems were bound to materialize. And once again, monitoring tools are emerging to help solve these issues.
Here are some of the top trends in container monitoring:
Container Challenges Abound
Kubernetes and containerization in general have grown so much that monitoring challenges abound. Cloud-based application management has become more complex, and blind spots are cropping up as the disposable nature of containers means that they introduce new layers of abstraction between the application and the underlying hardware that can be hard to track. Conventional monitoring tools weren’t designed to deal with this the increased volume of data generated by containers or the easy portability of so many interdependent components. A lot more telemetry data is needed to ensure observability into the performance and reliability of the containers.
SaaS Monitoring Solutions
The good news is that a number of SaaS solutions have appeared that aid with container monitoring and keep track of their ephemeral nature. These systems add greater observability, stronger visualizations of ever-changing container ecosystems, and what additional components have been added to systems, what happened when things go wrong, and more.
Al Brown, CTO of Veritone, noted that there are plenty of good SaaS platforms out there for the monitoring and recovery of container stacks. These include Dynatrace, Sysdig, Lacework, Datadog, SolarWinds, and Elasticsearch.
When a modest number of containers are in play, management is relatively easy. But once the numbers escalate, it can be hard to monitor or manage how many there are and what they are up to,
The vendor community and the open-source community courtesy of Kubernetes have responded with a variety of tools to give IT more assistance. IBM Turbonomic and Red Hat OpenShift provide automatic scaling capabilities to take care of configuration, scaling up and down, idling, reliability, and fault tolerance
“Auto scale is a great way to increase or decrease the desired count of containers automatically based on load or metrics,” said Brown.
Inevitably, better tools would arrive on the market to provide metrics to make monitoring of container numbers and behaviors easier. The open-source community, for example, has provided Prometheus as a metric monitoring solution, which is one facet of the and is a part of the Cloud Native Compute Foundation.
“Prometheus is being used more and now includes more business metrics,” said Brown. “This allows more than just monitoring of resources (CPU, memory, etc.) but application behavior and issues as well.”
Prometheus collects and stores a series of metrics information about containers and other parts of the IT landscape such web or application performance. These are stored with a timestamp of when they were recorded. The metrics themselves can be adjusted by IT to aid on specific aspects of container monitoring. Metrics might include request times, number of active connections, active queries, number of containers, etc.
Event-driven architectures are another way that container monitoring is being improved: Red Hat OpenShift, for example, helps IT to build loosely coupled and distributed apps connecting with a variety of built-in or third-party event sources or connectors powered by Operators.
“As a Knative distribution, it shares benefits such as container-based packaging format, scale-to-zero, sophisticated autoscaling mechanism based on HTTP consumption, production-grade support for event-driven serverless applications backed by Apache Kafka, and support for a function programming model,” said Naina Singh, principal product manager at Red Hat.