Frequently Asked Questions
Find answers to common questions about Inflight. Can't find what you're looking for? Contact our support team.
Getting Started
How long does it take to set up Inflight?
Initial setup typically takes about 10-15 minutes. You'll deploy the Inflight Agent as a sidecar to your services and connect at least one APM provider. The Agent automatically discovers your runtime configurations, and you'll see your first optimization recommendations within 24 hours once enough data has been gathered.
What do I need to install?
The Inflight Agent runs as a lightweight sidecar (~5MB) alongside your services in Kubernetes. It automatically discovers your runtime configurations (JVM flags, Go environment variables, resource limits) and registers with the Metrics Collector. Your APM credentials stay centralized—they're never deployed to your pods.
What if I use multiple APM providers?
Inflight supports connecting multiple APM providers simultaneously. This is common when different teams or environments use different tools. The Metrics Collector normalizes metrics from all sources into a unified view, so you get consistent insights regardless of which APM collected the data.
Platform & Architecture
What are the core components of Inflight?
Inflight consists of five core services: (1) Inflight Agent - lightweight sidecar for service discovery, (2) Metrics Collector - multi-APM aggregation with 3-tier storage, (3) Simulator - multi-fidelity prediction engine, (4) AI Advisor - intelligent optimization orchestration, and (5) Monitoring Service - centralized alert management.
What runtimes does Inflight support?
Inflight currently provides full support for six runtime environments: Java, Go, Rust, .NET, Python, and Elixir. Node.js support is coming soon. Inflight is designed to be runtime-agnostic, so adding new runtimes is a core capability of the platform.
How does the simulation engine work?
The Simulator uses your historical metrics to build models of your application's behavior. It operates in four fidelity modes—from fast statistical predictions to comprehensive discrete event simulation—automatically selecting the appropriate level based on change complexity and required confidence. All predictions are validated through backtesting against historical data.
How is this different from load testing or chaos engineering?
Load testing uses synthetic traffic in lower environments—it can't capture real production patterns. Chaos engineering intentionally disrupts services to test resilience. Inflight takes a different approach: it uses your actual production metrics (already being collected by your APM) to simulate changes offline. You get production-grade insights without any production risk or environment limitations.
What makes the AI Advisor different from other optimization tools?
The AI Advisor evaluates multiple optimization candidates simultaneously, not just a single recommendation. It uses Pareto frontier analysis to identify optimal trade-offs, validates every suggestion through simulation, and enforces policy gates based on your service tier and risk tolerance. Every recommendation is explainable with clear evidence.
Can Inflight automatically apply recommended changes?
We're building closed-loop deployment capabilities that will automatically implement approved recommendations that meet quality thresholds. This Deployment Manager will complete the optimization loop from observation through automated deployment. This feature is currently in development.
The Inflight Agent
What does the Inflight Agent discover?
The Agent automatically detects JVM configuration (GC algorithm, heap sizes, metaspace, flags), Go runtime settings (GOGC, GOMEMLIMIT, GOMAXPROCS), and Kubernetes resource specifications (CPU/memory requests and limits). This happens without any manual configuration.
How lightweight is the Agent?
The Agent is approximately 5MB with no external dependencies. It's designed to have minimal resource footprint and won't impact your application performance. It only reads configuration—it never modifies your running applications.
Why use an agent instead of just APM data?
APM providers report runtime metrics but often miss configuration details. The Agent captures the actual configuration your application is running with—GC settings, memory limits, thread pools—which is essential for accurate optimization recommendations. It also ensures configuration data stays current as your deployments change.
Data & Storage
How does the 3-tier storage work?
Metrics flow through three storage tiers optimized for different access patterns: Hot tier for real-time dashboards (hours to days retention), Warm tier for trend analysis and model calibration (weeks to months), and Cold tier for long-term archival (months to years). Data automatically moves between tiers based on age and access patterns.
What data does Inflight collect?
Inflight collects performance metrics from your APM providers—response times, throughput, error rates, memory usage, GC statistics—plus runtime configuration from the Agent. We do not collect application code, business data, or personally identifiable information.
How are my APM credentials stored?
All API credentials are encrypted at rest and in transit, stored only in the Metrics Collector—never in your application pods. Inflight uses read-only API access, meaning we cannot modify your APM configuration or access anything beyond metric data.
Can I self-host Inflight?
Enterprise customers can deploy Inflight in their own infrastructure for complete data sovereignty. Contact our sales team for details on self-hosted deployments.
Simulation & Safety
What are the different simulation fidelity modes?
The Simulator operates in four modes: STATISTICAL (fastest, for simple changes), HYBRID (balanced speed and accuracy), FULL (comprehensive discrete event simulation for complex changes), and DEGRADED (best-effort when data is limited). The system automatically selects the appropriate mode and can escalate to higher fidelity when confidence is low.
How accurate are the predictions?
Prediction accuracy depends on the quality and quantity of your metric data. With sufficient historical data, our models typically achieve high accuracy. Every prediction includes a confidence score, and the backtester continuously validates model accuracy against historical outcomes.
What safety verdicts can a simulation produce?
Simulations produce three possible verdicts: APPROVED (all safety thresholds met, safe to deploy), WARNING (some risk factors detected, validate in staging first), and REJECT (critical issues detected, change will not achieve intended outcome). Each verdict includes detailed reasoning.
What if a recommendation doesn't work as expected?
Each recommendation includes rollback guidance for reverting the change. We also recommend testing significant changes in staging environments first. The Monitoring Service tracks outcomes and will alert you if performance degrades after deployment.
Recommendations & Customization
Can I ignore or customize recommendations?
Absolutely. Inflight provides recommendations—you decide what to implement. You can dismiss recommendations that don't fit your needs, configure thresholds and preferences, and set policy gates based on service tier and risk tolerance.
Why do recommendations include evidence?
We believe in explainable AI. Every recommendation includes the reasoning behind it, the metrics that informed the suggestion, expected outcomes with confidence intervals, and trade-off analysis when applicable. This transparency helps you make informed decisions.
Can I customize the analysis reports?
Yes. Reports are dynamically generated, not static templates. You can create custom analysis perspectives for different stakeholders—detailed technical breakdowns for engineers, risk summaries for ops, or cost implications for leadership.
What types of optimizations does Inflight recommend?
Recommendations span memory optimization (heap sizing, GC tuning), performance tuning (thread pools, connection pools), stability improvements (OOM prevention, timeout optimization), and cost efficiency (right-sizing, over-provisioning detection).
Monitoring & Alerts
How does alert correlation work?
The Monitoring Service automatically groups related alerts using correlation keys, reducing noise by up to 90%. Multiple alerts for the same underlying issue become a single actionable item, and alerts are prioritized by impact and urgency.
What alert lifecycle states are there?
Alerts follow a clear lifecycle: OPEN (requires attention), ACKNOWLEDGED (being investigated), and CLOSED (resolved). Full audit trails track who worked on each alert, time-to-resolution, and resolution notes for future reference.
Can I integrate alerts with my existing tools?
Yes. Alerts can be configured to notify via email, Slack, PagerDuty, or webhook. The Monitoring Service aggregates alerts from all Inflight components into a single unified view.
Integration & Workflow
Does Inflight integrate with CI/CD pipelines?
Yes. You can use Inflight's API to validate configuration changes as part of your deployment pipeline. If a proposed change doesn't pass simulation safety checks, your pipeline can be configured to block the deployment.
Can multiple team members use Inflight?
Yes. Inflight supports teams with role-based access control. You can have different permissions for viewing recommendations, approving changes, and managing integrations.
Which APM providers are supported?
Inflight integrates with Dynatrace, Datadog, New Relic, and SigNoz. Additional providers can be added based on customer requirements. The Metrics Collector normalizes data from all providers into a consistent format.
Pricing & Plans
Is there a free trial?
Yes. We offer a free trial so you can experience Inflight with your own services before committing. Contact us to get started.
How is pricing calculated?
Pricing is based on the number of services you're optimizing and your data retention requirements. We offer Starter, Professional, and Enterprise tiers to fit teams of different sizes. Contact us for detailed pricing information.
What's included in Enterprise?
Enterprise includes advanced features like self-hosted deployment options, custom model training, extended data retention across all storage tiers, priority support, SLA guarantees, and dedicated success management. It's designed for organizations with complex requirements or compliance needs.
Still Have Questions?
Our team is happy to help with any questions not covered here.