Cloud computing has revolutionised how we conceive, deploy, and manage applications. But this revolution comes at a cost: ever-increasing complexity. Manually managing thousands of services, virtual machines, and containers has become a colossal challenge, a source of errors and slowdowns.
What if infrastructure could manage itself? This is the promise of the autonomous cloud, a fascinating field of research at the intersection of cloud computing and autonomous systems.
The Convergence of Two Worlds
On one hand, cloud computing offers virtually infinite computing and storage power that is flexible and distributed. On the other hand, autonomous systems are entities capable of making decisions and acting without human intervention to achieve specific objectives and adapt to a changing environment.
The autonomous cloud merges these two ideas: it aims to create a cloud infrastructure endowed with the capabilities of an autonomous system.
The Ultimate Example of an Autonomous System: The Human Body
To understand this concept, we need not look far. The most sophisticated autonomous system we know is the human body. Think about it:
Self-healing: When you cut yourself, your body triggers a complex process of coagulation and healing without you having to think about it.
Self-protecting: Your immune system continuously detects and neutralises threats (viruses, bacteria) to keep you healthy.
Self-optimising: When you exercise, your body adapts by strengthening your muscles and improving your cardiovascular capacity.
The autonomous cloud aspires to reach this level of intrinsic intelligence, where infrastructure can repair, protect, and optimise itself completely independently.
From Biology to Technology: The Mars Rover
Another emblematic example is that of space rovers, like Perseverance. Due to the several-minute communication delay between Earth and Mars, the rover cannot be piloted in real-time. It must make critical decisions alone: circumvent an unexpected rock, manage its energy levels, or analyse an interesting stone. It operates autonomously to fulfil its mission in a hostile and distant environment.
Now that we’ve established the conceptual foundations, how do we build such a system? What is the “brain” that enables this autonomy?
The Architecture of the Autonomous Cloud – At the Heart of the MAPE-K Loop
In the first part, we defined the autonomous cloud by drawing inspiration from systems as complex as the human body and Martian rovers. But to move from concept to reality, we need a plan, an architecture.
This is where the MAPE-K loop comes in. IBM proposed this reference model, which has become a de facto standard for designing autonomous systems. This loop is the true engine of the system’s intelligence.
MAPE-K is the acronym for Monitor, Analyse, Plan, Execute, all orchestrated by a common knowledge base, the Knowledge (K). Let’s break down this cycle.
The Four Phases of the Autonomous Cycle
Monitor: The Eyes and Ears of the System
The first step is to collect raw data on the system’s state and environment. This includes performance metrics (CPU, RAM), logs, execution traces, security alerts, etc. This is the perception phase.
Analyse: Making Sense of the Data
Raw data is not enough. In this phase, the system analyses the collected information to detect patterns, correlations, or anomalies. Is an increase in response time related to database overload? Is a security alert a real threat? This is where information becomes valuable.
Plan: Deciding on the Course of Action
Once the problem is identified and understood, the system must develop an action plan to achieve the desired state. If an application is slowing down, the plan could be to start two new servers, add them to the load balancer, then verify that performance has returned to normal. This phase transforms understanding into intention.
Execute: Acting on the World
The final step consists of implementing the plan. The system interacts with the underlying infrastructure (hypervisor, container orchestrator, firewall) to execute the planned actions. Commands are sent, and the cycle begins again at the “Monitor” phase to observe the effects of the changes.
The “K”: The Crucial Role of Knowledge
The MAPE loop would be nothing without the “K.” This knowledge base centralises all the information necessary for decision-making: policies (e.g., “response time must never exceed 200 ms”), historical data, system topology, business objectives, etc. It is the system’s memory and long-term intelligence, consulted and updated at each step of the cycle.
This architecture provides the perfect framework for building an autonomous cloud system.
Our Autonomous Ecosystem – LabelIT, ObservIT, and StructurIT
After exploring the concept of the autonomous cloud and its reference architecture, the MAPE-K loop, it’s time to present our concrete approach. Our vision is embodied through an ecosystem of three complementary projects, each playing a specific but interdependent role in achieving complete autonomous management.
These three pillars are: LabelIT, ObservIT, and StructurIT. Together, they form a coherent and powerful implementation of the MAPE-K loop.
ObservIT: Perception (Monitor & Analyse)
ObservIT is the project that enables us to overcome the barriers related to observability. It constitutes our system’s senses, with the mission of continuously monitoring the infrastructure by collecting a wide range of telemetric data. But its task doesn’t stop there; it is also responsible for the first phase of analysis. By aggregating, correlating, and filtering data, ObservIT transforms informational noise into relevant signals. It detects symptoms: increasing latency, filling disk space, or suspicious network activity.
Role in MAPE-K: Monitor and part of Analyse.
LabelIT: Context and Knowledge (Knowledge)
A symptom alone is not enough. Is an increase in CPU usage critical? It depends. If it’s a production server, the answer is yes. If it’s a test environment, it’s less urgent. This is where LabelIT comes into play. Its function is to enrich raw data with business and technical context. It “labels” resources: “database-production”, “web-server-clientA”, “accounting-batch”.
LabelIT doesn’t just label, it builds and maintains the knowledge base (K).
Role in MAPE-K: Feeds and structures the Knowledge.
StructurIT: Decision and Action (Plan & Execute)
Armed with ObservIT’s observations and the context provided by LabelIT, StructurIT represents the most advanced phase of our loop: transforming intention into technical reality. Its role is not simply to react, but to build.
The StructurIT project aims to automate the design of technical architectures using generative artificial intelligence. It takes as input a business need expressed in natural language (for example: “I want a secure architecture to store customer data”), understands it, reasons about constraints (cost, performance, security), and then plans the ideal infrastructure.
Finally, it executes this plan by directly generating Infrastructure as Code (IaC), like Terraform, ready to be deployed. It doesn’t just choose an action from a list, it creates the solution.
Role in MAPE-K: Plan and Execute, relying on generative AI to create complex plans and translate them into concrete actions (code).
Conclusion: Towards a Partner Infrastructure
The autonomous cloud, far from being a simple futuristic concept, represents the next logical evolution of cloud computing. We have moved from physical servers to virtual machines, then to containers, each step bringing more flexibility but also a new layer of complexity. Autonomy is not simple automation; it’s a paradigm shift. It’s no longer about giving orders to a passive infrastructure, but collaborating with an intelligent and proactive system.
By drawing inspiration from the resilience of the human body and the ingenuity of Martian rovers, and by relying on a robust framework like the MAPE-K loop, we have laid the foundations of an infrastructure capable of perceiving itself, understanding itself, and acting. The true power of this approach lies in the synergy between its components. Observation without context is blind, and action without perception is dangerous.
This conviction gave birth to our ecosystem. ObservIT, LabelIT, and StructurIT are not isolated tools but the interdependent organs of this digital brain. Together, they create a virtuous circle where infrastructure learns, enriches itself, and becomes more relevant with each cycle, transforming the promise of the autonomous cloud into operational reality.
In the coming weeks, we will publish more detailed articles focusing on StructurIT, ObservIT, as well as LabelIT.