This article is based on a perspective piece originally published on finextra.com in early 2020. The trends described reflect the DevOps community’s direction at that time and remain highly relevant today.
As cloud-native adoption accelerated into 2020, a common theme emerged across DevOps teams worldwide: how do you deliver reliable services at scale without proportionally scaling your headcount? Four trends dominated that conversation — Site Reliability Engineering, Kubernetes adoption, service meshes, and AIOps. Understanding each one gives engineering teams a practical roadmap for building resilient, scalable systems.
1. Site Reliability Engineering (SRE)
As organisations moved more workloads to the cloud, a critical operational question surfaced: how do you support millions of global users without running a large, round-the-clock operations team? The answer many organisations turned to was Site Reliability Engineering (SRE).
At its core, SRE pairs software development skill with operational responsibility. The goal is to build services that can scale without requiring a linear increase in the human effort needed to maintain them. SRE teams achieve this by eliminating toil — repetitive, manual operational work — and by encoding reliability directly into systems through software and automation.
When failures occur, SREs conduct blameless post-mortems, analysing what happened before, during, and after an incident to identify root causes and reduce the risk of recurrence. This forensic approach borrows from industries such as aviation and healthcare, where structured incident review is standard practice. The output is not just a report — it feeds directly back into the service as software improvements.
By 2020, a clear trend emerged: Site Reliability Engineers were being embedded within or alongside engineering teams rather than sitting in a separate, centralised operations function. In this model, the team that builds a service also becomes responsible for its availability — a natural fit for DevOps-oriented organisations.
One important caveat: what works at Google, Netflix, or Spotify may not translate directly to smaller organisations or those with different engineering cultures. The definition and scope of an SRE role remains fluid, and finding the right implementation for a given organisation’s size and culture is an ongoing challenge.
2. Kubernetes Adoption
Kubernetes adoption continued its strong growth trajectory into 2020, driven by the widespread uptake of containerised microservices. A notable shift in mindset accompanied this growth: rather than retrofitting Kubernetes onto existing architectures after the fact, more engineering teams were designing software with Kubernetes as the target platform from the start.
In earlier eras of open-source development, Linux served as the foundational platform. Kubernetes was beginning to occupy a similar role for cloud-native software: a common substrate on which solutions are built and distributed.
Several platform capabilities supported this maturation:
- Serverless workloads via Knative: A major discussion point at KubeCon, Knative allows teams to run event-driven, serverless workloads on top of Kubernetes without managing the underlying infrastructure.
- Kubernetes Operators: Operators give software vendors a standardised way to package and deliver their applications on the Kubernetes platform, accelerating the growth of a cloud-native software marketplace.
- Commercial Kubernetes distributions: Platforms from major vendors continued to mature, offering improved developer experience, faster paths to production readiness, and simplified IT maintenance.
- Federation features: There was growing expectation that federation capabilities would improve coordination across multiple Kubernetes clusters, enabling more robust multi-cloud and distributed deployments.
The large and active Kubernetes community was also expected to accelerate the pace of feature releases, further cementing the platform’s central role in cloud-native infrastructure.
3. Service Meshes
Microservices architecture offers significant benefits — portability, independent deployability, and team autonomy — but it introduces real operational complexity. As organisations moved from monolithic applications to dozens or hundreds of microservices spread across hybrid and multi-cloud environments, the challenge of managing service-to-service communication grew substantially.
Service meshes emerged as a direct response to this complexity. A service mesh sits as a dedicated infrastructure layer that controls and observes network traffic between microservices. Rather than baking networking concerns into application code, the mesh handles them transparently — typically using sidecar proxies deployed alongside each service instance.
For DevOps and SRE teams, a service mesh provides practical tools for complex operational requirements, including:
- Traffic management: Canary rollouts, A/B testing, and traffic shifting between service versions.
- Observability: Distributed tracing, metrics, and a unified view of service health across the entire mesh.
- Security: Mutual TLS (mTLS) for end-to-end encryption and fine-grained access control between services.
- Resilience: Automatic retries, circuit breaking, and timeout enforcement without code changes.
As more enterprises undertook the journey from monolith to microservices, service meshes were increasingly recognised as a necessary part of the operational toolkit rather than an optional add-on. The number of competing offerings — including Istio, Linkerd, and Consul Connect — was expected to grow throughout 2020.
4. AIOps
Modern distributed systems generate enormous volumes of telemetry: metrics, logs, traces, and alerts from dozens of monitoring tools simultaneously. Human operators are fundamentally ill-suited to manually correlating this volume of data in real time, particularly during high-pressure incidents when speed of response is critical.
AIOps — the application of artificial intelligence and machine learning to IT operations — addresses this problem directly. Rather than presenting operators with a flood of raw alerts (analogous to every smoke alarm in a building triggering at once), AIOps platforms ingest data from multiple sources, correlate anomalies, suppress noise, and surface actionable insights.
Key capabilities associated with AIOps platforms include:
- Anomaly detection: Identifying deviations from baseline behaviour across metrics and logs before they escalate into incidents.
- Alert correlation: Grouping related alerts from multiple monitoring tools into a single, coherent incident view.
- Predictive analytics: Using historical patterns to predict where and when failures are likely to occur, enabling proactive remediation.
- Noise reduction: Filtering out low-priority or duplicate alerts so on-call engineers can focus on what actually matters.
The synergy between AIOps and SRE practice is natural: both are focused on improving the speed and quality of incident detection and response. As DevOps and SRE adoption drove continuous improvement in incident management processes, interest in AI-assisted operations tooling accelerated. By 2020, dedicated AIOps platforms were proliferating, and major observability and monitoring vendors were integrating AIOps capabilities into their existing products.
The Bigger Picture
Across all four trends, a single overarching theme connects them: the challenge of delivering reliable services at scale in the cloud with sustainable team structures. SRE addresses the organisational and cultural model. Kubernetes provides the platform. Service meshes manage the complexity of distributed communication. AIOps handles the observability and incident response challenge at machine scale.
Taken together, these four trends represent not just a snapshot of 2020 but a durable direction for the decade — one that engineering teams continue to navigate today.