IBM Integrates iter8 With Kiali Interface for Istio
IBM and its Red Hat subsidiary have combined their respective expertise to make it easier to deploy microservices across the Istio service mesh using iter8, an open source project IBM created that employs machine learning algorithms to optimize rollouts of microservices-based applications.
As part of that effort, iter8 has been integrated with Kiali, an open source project initiated by Red Hat that provides a user interface for managing the Istio service mesh. IT teams via that interface can validate Istio configuration in addition to visualizing metrics generated by specific microservices and connectivity patterns between microservices.
Fabio Oliveira, a research scientist and manager at IBM’s Thomas J. Watson Research Center, says the goal is to make it easier to both deploy Istio services and then manage the overall IT environment.
Most organizations don’t deploy a service mesh until they know they will have hundreds of microservices to manage. Prior to that level, proxy software or an Ingress controller is generally sufficient. IBM developed Istio in collaboration with Google and Lyft and is now bringing other open source initiatives to bear as part of an effort to foster adoption of Istio and other cloud-native platforms.
IBM expects to be able to apply the observability capabilities provided by iter8 across a wide range of those cloud-native platforms, he says.
While most IT teams historically have monitored individual components of an IT environment, there’s been a marked shift toward observability platforms that promise to provide more context. Those platforms achieve that goal by aggregating metrics from multiple infrastructure and application sources. However, those metrics require advanced analytics infused with machine learning algorithms to provide actionable insights across increasingly complex distributed computing environments.
It’s not clear just yet which service mesh will prevail in Kubernetes environments, but the level of support Istio enjoys today suggests it will have significant staying power. The challenge IT teams will encounter is that while Istio makes it possible to better manage microservices, it doesn’t provide any core analytics capability that would enable IT teams to optimally manage those microservices.
Oliveira says iter8 makes it easier for IT teams to discover trends spanning multiple versions of microservices, in addition to tracking canary releases of microservices based on their behavior and the service level objectives defined by the IT organization. During a canary rollout experiment, iter8 treats the current version of a microservice and the canary version as rivals. It then periodically assesses the canary’s quality and adjusts how traffic is split between the two competing microservices. As more data becomes available, iter8 will gain confidence in its ability to declare whether the canary will succeed or fail. It will then gradually shift traffic toward either the current version or the canary based on that assessment.
One of the best things about microservices-based applications is they are designed to degrade gracefully versus outright crash. The trouble is, given all the dependencies that exist between microservices, it’s often difficult to determine the root cause of any degradation. The longer-term goal, of course, should be to prevent that degradation from ever happing in the first place.
