Opstral glossary
Plain-English definitions of the platform terms used across this site, plus the AIOps industry terms behind them. If you have seen a word like Sentinel, ProcBot, Sherlock, OIAO, or MOP and wondered what it means, start here.
Opstral is an autonomous AIOps platform. The terms below describe its agents, its operating method, and the industry concepts it builds on.
79 terms
Platform & agents
- Opstral
- An autonomous AIOps platform that unifies operations across incident management, security, data, cloud cost, and business processes, under one intelligence layer powered by Sentinel AI. It is operated by VWAVES Technologies Pvt. Ltd.
- Sentinel(Sentinel AI)
- Opstral's central reasoning engine. Sentinel autonomously observes signals across every connected tool, correlates them, identifies the root cause of an incident, and orchestrates the resolution. It is the intelligence core that the rest of the platform, including ProcBot and Sherlock, runs on.
- ProcBot
- Opstral's execution agent. ProcBot runs validated MOPs (playbooks) using Ansible and shell commands to remediate incidents, either fully autonomously when policy allows or after a human approves the action.
- Sherlock
- Opstral's post-incident fix validation engine. After a fix is applied, Sherlock confirms the incident is genuinely resolved, scores how effective the procedure was, detects recurring issues, and feeds those learnings back to Sentinel so the system keeps improving.
- Orchestrator
- Opstral's cross-platform action orchestration engine. It federates detection and execution across ITSM, ERP, CRM, HR and identity platforms through agent-to-agent coordination, so an incident that spans systems gets one owner and one resolution path.
- Copilot
- Conversational operations. Copilot lets any engineer investigate in plain English across every connected source, so L1 resolves what used to escalate to L2.
- Custom Agent Studio
- The build surface for custom agents and MOPs. Teams extend the platform with their own procedures and automations, under the same governance as everything else.
The OIAO method
- OIAO
- Opstral's closed-loop operations cycle: Observe, Investigate, Act, Optimize. Every incident flows through the same four phases, and the Optimize phase feeds its learnings back into Observe, so signal correlation gets sharper over time.
- Observe, Investigate, Act, Optimize
- The four phases of the OIAO loop. Observe ingests and normalizes signals (metrics, logs, traces, alerts, security events). Investigate performs automated root-cause analysis, maps blast radius, and selects the right MOP. Act executes the fix through ProcBot or routes it for human approval. Optimize validates the outcome through Sherlock and scores the procedure.
- MOP(Method of Procedure)
- A validated runbook that codifies exactly how to resolve a specific type of incident. Opstral stores MOPs in a library, ProcBot executes them, and Sherlock scores their effectiveness so the best procedures rise to the top.
- Action Ticket
- The governed unit of execution in Opstral. Every production-changing action runs through an Action Ticket: pre-checked, approval-gated where policy requires, reversible, and fully audited.
- Graduated autonomy
- The trust model for automation. The platform starts advisory, moves to approval-gated execution, and earns fully autonomous operation per procedure and per system as confidence is proven against live results.
Core components
- Integration Connectors
- The connectivity layer: 2,000+ governed connectors that bring signals in from the systems you already run and carry approved actions back out.
- Data Governance Fabric
- The data control layer. It curates and anonymises data before it reaches any model and governs it end to end, with lineage and policy-based access.
- Alerts & Rule Engine
- The deterministic layer. Where an outcome must be exact, or a regulator expects it, rules run instead of inference, so the known and the regulated are handled with certainty.
- Data Ingestion & Transformation
- The ingestion layer: batch, streaming and change-data-capture at big-data scale, shaping data before it reaches the platform.
Operational pillars
- Telemetry Ops
- OpenTelemetry-native telemetry at petabyte scale: metrics, logs and traces ingested, correlated and queryable in one place, feeding every investigation Sentinel runs.
- Service Ops
- End-to-end application observability, tracing a request from the user all the way to the database across every service in between.
- Infra Ops
- Infrastructure operations: cluster and workload health, lifecycle management, resource management, and automated remediation.
- Security Ops
- Security operations and threat detection, with detections mapped to MITRE ATT&CK techniques and tactics.
- Data Ops
- Data pipeline operations: job health monitoring, data-quality drift detection, and ingestion lag tracking.
- Fin Ops
- Cloud cost intelligence: real-time spend monitoring, cost anomaly detection, and rightsizing recommendations.
- Process Ops
- Process operations: discovering, monitoring and orchestrating how work flows across systems, end to end.
- AI Ops
- Operations for models and AI pipelines: model and pipeline monitoring, drift and data-quality checks, and inference performance and cost.
- DevSec Ops
- Security and delivery in one loop: pipeline and build health, vulnerability and policy checks, and change risk scoring with rollback.
- Managed Ops(MSOps)
- Runs the managed application estate as one control surface: incident command with blast-radius impact, service topology, SLA and error-budget governance, and governed service-to-process mapping. Related to Managed Services.
Industry & technical terms
- AIOps
- Artificial Intelligence for IT Operations. Applying AI and machine learning to automate and improve IT operations such as monitoring, incident response, and root-cause analysis.
- Managed Services(application estate operations)
- Ongoing managed operations for an enterprise's application estate. Federated managed services refers to orchestrating application management across multiple providers or business units from a single intelligence layer.
- MTTR(Mean Time To Resolution)
- The average time taken to resolve an incident, from detection to confirmed fix. A core measure of operations performance.
- RCA(Root Cause Analysis)
- Identifying the underlying cause of an incident rather than just its symptoms, so the same issue does not recur.
- SIEM(Security Information and Event Management)
- Tooling that aggregates and analyzes security events from across an environment to detect threats.
- ITSM(IT Service Management)
- The practice and tooling for managing IT services, including incident ticketing and change management.
- Runbook
- A documented, repeatable procedure for handling a specific operational task or incident. In Opstral, runbooks are formalized as MOPs.
- Blast radius
- The scope of systems, services, or users affected by an incident. Mapping blast radius is part of the Investigate phase.
- A2A orchestration(agent-to-agent)
- Multiple autonomous agents, Sentinel, ProcBot, and Sherlock, coordinating across systems to resolve an incident end to end without a human stitching the steps together.
- MTTD(Mean Time to Detect)
- Mean time to detect: how long it takes to notice an incident is happening. Continuous correlation and anomaly detection push MTTD down for known patterns.
- MTBF(Mean Time Between Failures)
- Mean time between failures: the average interval between incidents. A stability measure that should lengthen as Optimize-phase learnings compound.
- Rollback
- Reverting a change to its previous state. Every Action Ticket carries a rollback path, armed before the change is made, not improvised after.
- NOC(Network Operations Center)
- Network operations center: the team and toolset watching production around the clock. Opstral is built by a team that runs carrier-scale NOC operations.
Governance, security & compliance
- RBAC(Role-Based Access Control)
- Role-based access control: access granted based on assigned roles. Applied across every Opstral console, API and action from day one.
- ABAC(Attribute-Based Access Control)
- Attribute-based access control: fine-grained, context-aware permissions that consider attributes of the user, the resource and the situation, not just the job title.
- Audit trail
- The continuous record of every decision, action, approval and rollback, generated as operations run, so compliance evidence exists the moment it is asked for.
- Data lineage
- The traceable map of where data originated and every transformation it passed through, maintained automatically for explainability and audit.
- PII(Personally Identifiable Information)
- Personally identifiable information: data that can identify an individual. The Data Governance Fabric masks and anonymises PII before any data reaches a model.
- MITRE ATT&CK
- An industry knowledge base of adversary tactics and techniques. Security Ops detections are mapped to ATT&CK so triage starts with context, not a bare alert.
- SOAR(Security Orchestration, Automation and Response)
- Security orchestration, automation and response. Opstral ships purpose-built SOAR connectors so security response runs under the same governed execution as the rest of operations.
- Air-gapped deployment
- Running the full platform inside a network with no internet connectivity and no data egress. Opstral operates air-gapped for sovereign and regulated environments.
- GDPR(General Data Protection Regulation)
- The EU's General Data Protection Regulation. Continuous audit trails, lineage and PII controls support GDPR reporting from live operations rather than manual evidence hunts.
- HIPAA(Health Insurance Portability and Accountability Act)
- A US healthcare data regulation. The platform's data governance and audit capabilities support HIPAA-aligned handling of operational data.
- SOX(Sarbanes-Oxley Act)
- A US financial reporting regulation. Continuous, cross-system audit trails support SOX compliance evidence as a report, not a project.
Integration & standards
- OpenTelemetry
- An open observability standard for metrics, logs and traces. Opstral is OpenTelemetry-native, so your telemetry stays portable and vendor lock-in stays out.
- CDC(Change Data Capture)
- Change data capture: replicating database changes in near-real time. One of the ingestion modes supported by Data Ingestion & Transformation, alongside batch and streaming.
- Kubernetes
- An open-source container orchestration system. Opstral deploys and scales its workloads on Kubernetes, on-premises or in your cloud.
- Apache Kafka
- A distributed event-streaming platform, supported as a streaming signal source and sink through Integration Connectors.
- Webhooks
- HTTP callbacks that push events in real time. Supported both as an input signal source and as an output channel for downstream systems.
- REST API(Representational State Transfer)
- The standard web interface used across connectors and for programmatic access to platform capabilities.
AIOps concepts
- Alert noise
- The high volume of low-value, duplicate, or non-actionable alerts operations teams receive. Industry estimates suggest only a small fraction of alerts require action; cutting alert noise is a core job of an AIOps platform.
- Alert fatigue
- The desensitisation that sets in when engineers are flooded with so many alerts that genuine incidents get missed or ignored. A direct consequence of alert noise.
- Event correlation
- Grouping related signals (alerts, events, metrics) from different tools into a single incident, so teams see one problem with full context instead of hundreds of disconnected fragments.
- Deduplication
- Collapsing repeated or identical alerts for the same underlying issue into one, so responders are not paged many times for a single problem.
- Enrichment
- Adding context to a raw alert, such as the owning service, recent changes, and topology, so a responder can act without hunting across tools for information.
- Anomaly detection
- Using statistical or machine-learning methods to flag behaviour that deviates from a normal baseline, often surfacing a problem before it becomes a hard failure.
- Predictive alerting
- Warning of a likely incident ahead of time, based on early signals and learned patterns, rather than only reacting after an outage has already happened.
- Observability
- The ability to understand a system's internal state from the data it emits, typically metrics, events, logs, and traces. Opstral sits on top of your observability stack rather than replacing it.
- Telemetry(MELT)
- The metrics, events, logs, and traces (MELT) that systems emit and that monitoring and AIOps tools consume to understand health and diagnose incidents.
- Incident
- An unplanned disruption or degradation of a service that requires a response. The unit of work an AIOps platform is built to detect, investigate, and resolve.
- Severity(SEV)
- A rating of an incident's business impact, for example SEV1 for a critical outage down to SEV3 or SEV4 for minor issues. Severity drives how an incident is escalated and who responds.
- Escalation policy
- The rules that define who is notified, and in what order, if an incident is not acknowledged or resolved within a set time.
- On-call
- The rotation of engineers responsible for responding to incidents outside normal working hours. Reducing on-call burden is a primary goal of autonomous operations.
- Postmortem(post-incident review)
- A blameless review after an incident that captures what happened, the root cause, and the actions needed to prevent recurrence.
- SLO(Service Level Objective)
- A target for service reliability, for example 99.9% availability, that a team commits to and measures against.
- SLA(Service Level Agreement)
- A contractual commitment on service performance between a provider and a customer, often carrying penalties for breach.
- Error budget
- The allowable amount of unreliability, calculated as one minus the SLO, that a team can spend before reliability work takes priority over shipping new features.
- Self-healing
- Operations that detect and remediate issues automatically, without human intervention. Opstral's closed loop, Sentinel investigating, ProcBot acting, and Sherlock validating, is a self-healing pattern.
- Federated detection
- Correlating signals across platform boundaries, ITSM, ERP, CRM, HR and identity, to find root causes no single-platform tool can see. Performed by Sentinel through the Orchestrator.
AI concepts
- Agentic AI
- AI that can autonomously make decisions, plan a sequence of actions, and pursue a goal with minimal human intervention, adapting as conditions change. Sentinel, ProcBot, and Sherlock are agentic.
- GenAI(Generative AI)
- AI that creates new content, such as text, code, or summaries, from patterns in its training data. In operations it powers plain-English investigation and automatically written incident summaries.
- LLM(Large Language Model)
- An AI model trained on large amounts of text that can understand and generate human-like language. LLMs are what let engineers query operations in plain English.
- Human-in-the-loop
- Keeping a human in the decision path to approve or guide an autonomous action. Opstral acts autonomously when policy allows and routes to a human when judgment is required.