The Complete Architect’s Decision Guide (2026)

Microsoft Sentinel provides three distinct storage options:

  1. Analytics Tier
  2. Sentinel Data Lake
  3. Data Archive

Microsoft documentation explains what each option is, but it does not provide a clear, operational framework for deciding where each log or table should live in real-world Sentinel environments.

Last verified against Microsoft documentation: Feb 2026


Why Storage Tiering Is the Most Important Sentinel Decision

Incorrect storage decisions lead to:

  • High ingestion and retention costs
  • Broken or inefficient detections
  • Poor investigation experience
  • Slow incident response
  • Unscalable Sentinel deployments

⚠️ Most Sentinel problems are architecture problems — not analytics rule problems.


First Principles: Understand the Three Storage Tiers

Before making any decision, we must clearly understand what Microsoft designed each tier to do — and what it is NOT designed to do.


1️⃣ Analytics Tier (Log Analytics Workspace)

Purpose (Microsoft intent)

The Analytics Tier is designed for:

  • Near-real-time detections
  • Scheduled and NRT analytics rules
  • UEBA and Fusion correlation
  • Incident creation
  • Automation rules and playbooks

Benefits

  • Low-latency KQL execution
  • Full KQL language support
  • Mandatory for analytics rules
  • Required for SOC operations
  • Native Defender and automation integration

Limitations

  • High ingestion cost
  • Expensive long-term retention
  • Not designed for multi-year data
  • Cost grows rapidly with volume

⚠️ KQL Limitations (Explicit)

  • Optimized for frequent, operational queries, not massive historical scans
  • Complex joins across high-volume tables can be expensive
  • Cannot query archived data without rehydration
  • Not suitable for very large, investigation-heavy datasets

Best suited for

  • Detection-driven logs
  • UEBA and correlation inputs
  • SOC dashboards and automation

📘 Microsoft references
https://learn.microsoft.com/azure/sentinel/overview
https://learn.microsoft.com/azure/sentinel/analytics-rule-concepts


2️⃣ Sentinel Data Lake

Purpose (Microsoft intent)

The Sentinel Data Lake is designed for:

  • Long-term security data retention
  • Large-scale investigations
  • Threat hunting over months or years
  • Retroactive detection scenarios
  • Forensics and research

Benefits

  • Significantly cheaper than Analytics for long retention
  • Scales to massive datasets
  • KQL supported for investigation and hunting
  • Ideal for forensic timelines and historical analysis

Limitations

  • Not designed for real-time alerting
  • Cannot directly trigger analytics rules
  • Higher query latency than Analytics
  • Not suitable for SOC dashboards

⚠️ KQL Limitations (Explicit)

  • KQL is supported, but optimized for investigative workloads
  • No direct alerting or scheduled analytics rules
  • Queries are slower than Analytics Tier
  • Requires architectural planning for promotion or mirroring

🔄 Important Clarification: Indirect Detection Support

While Data Lake cannot directly trigger analytics rules, it can be used indirectly:

  • KQL jobs can run on Data Lake data
  • Results can be promoted or mirrored into Analytics Tier
  • Promoted data can then be used for detections or alerts

This enables hybrid use cases, such as:

  • Retroactive detection
  • IOC backtracking
  • Historical anomaly discovery

📘 Microsoft references
https://learn.microsoft.com/azure/sentinel/datalake/sentinel-lake-overview
https://learn.microsoft.com/azure/sentinel/manage-data-overview


3️⃣ Data Archive (Compliance-Only Tier)

Purpose (Microsoft intent)

Data Archive exists to support:

  • Regulatory retention
  • Legal and audit requirements
  • Long-term storage with minimal access

Benefits

  • Lowest storage cost
  • Ideal for multi-year compliance retention
  • Designed for “retain but rarely access” data

Limitations

  • Not usable for detections
  • Not usable for investigations
  • Requires rehydration before access
  • High access latency

⚠️ KQL Limitations (Explicit)

  • KQL cannot run directly on archived data
  • Data must be rehydrated into Analytics Tier
  • Not suitable for hunting, dashboards, or automation

Best suited for

  • Compliance-only logs
  • Legal hold scenarios
  • Rarely accessed audit data

📘 Microsoft references
https://learn.microsoft.com/azure/azure-monitor/logs/logs-data-archive
https://learn.microsoft.com/azure/azure-monitor/logs/logs-data-retention


Cross-Tier Querying Clarification (Important Nuance)

  • Analytics Tier → Data Archive
    ❌ Not possible without rehydration
  • Analytics Tier ↔ Sentinel Data Lake
    ✅ Possible via mirroring / promotion for the analytics retention window (commonly 30–90 days)

This allows:

  • Short-term cross-tier analysis
  • Controlled detection enablement
  • Cost-efficient hybrid designs

Side-by-Side Comparison

DimensionAnalytics TierSentinel Data LakeData Archive
Primary purposeDetection & responseInvestigation & huntingCompliance retention
Alerting✅ Yes❌ No (indirect only)❌ No
UEBA / Fusion✅ Yes❌ No❌ No
KQL supportFullSupported (investigative)❌ Rehydrate only
Query frequencyHighMedium / LowVery rare
LatencyLowMediumHigh
Retention costHighMedium / LowLowest
Best forActionUnderstandingObligation

Pre-Step Before Tiering: Consider Basic Logs

Before deciding Analytics vs Data Lake vs Archive, evaluate Basic Logs for noisy sources.

Basic Logs (Context)

  • Lower ingestion cost
  • Limited KQL support
  • No analytics rules
  • Suitable for verbose logs (DNS, proxy, firewall)

📘 Microsoft reference
https://learn.microsoft.com/azure/azure-monitor/logs/logs-basic

🧠 Architect tip:
First decide Standard vs Basic, then decide Analytics vs Data Lake vs Archive.


The Core Architect Rule (Non-Negotiable)

If a Sentinel analytics rule, UEBA model, or Fusion correlation depends on a table, that table MUST be in Analytics Tier.

No exceptions.


The Complete Sentinel Storage Decision Framework

Use this logic for every log source or Sentinel table.

Step 1: Is the data used by analytics rules, UEBA, or Fusion?

→ Analytics Tier

Step 2: Does it trigger incidents or automation?

→ Analytics Tier

Step 3: Is it needed for investigations or threat hunting?

→ Sentinel Data Lake

Step 4: Is it required only for compliance or audit?

→ Data Archive

Step 5: Is it noisy with limited security value?

→ Basic Logs or do not ingest


Sentinel Tables — Practical Placement Examples

Identity

  • SigninLogs → Analytics + Data Lake
  • AuditLogs → Analytics + Data Lake

Endpoint

  • DeviceLogonEvents → Analytics
  • DeviceNetworkEvents → Filtered Analytics + Data Lake

Network

  • CommonSecurityLog → Filtered Analytics + Data Lake
  • Raw DNS / proxy logs → Basic Logs or Data Lake

📘 Reference
https://learn.microsoft.com/azure/sentinel/connect-data-sources


Cost & Pricing (Microsoft-Aligned, Simplified)

📘 Official pricing
https://azure.microsoft.com/en-us/pricing/details/microsoft-sentinel/

Example (Approximate)

  • Analytics Tier
    ~$2.50 per GB ingested + retention costs
  • Sentinel Data Lake
    ~$0.01 per GB per month beyond Analytics retention
  • Data Archive
    Lowest cost storage + rehydration charges when accessed

💰 Architect rule:
If a log increases cost but does not improve detection, it does not belong in Analytics Tier.


Common Architecture Mistakes (Real-World)

❌ Using Analytics Tier as a data dump
❌ Using Data Lake for direct detections
❌ Using Data Archive for investigations
❌ Ignoring KQL limitations per tier
❌ Skipping Basic Logs for noisy sources


Sentinel Design Checklist (Printable)

Architecture

  • ⬜ Detection tables identified
  • ⬜ Investigation tables separated
  • ⬜ Compliance-only data archived

Ingestion

  • ⬜ Basic Logs evaluated
  • ⬜ DCRs filtering noise
  • ⬜ ASIM normalization applied

Detection

  • ⬜ Analytics rules only query Analytics Tier
  • ⬜ Indirect Data Lake promotion defined

Cost

  • ⬜ High-volume tables controlled
  • ⬜ Retention optimized per tier

📘 Best practices
https://learn.microsoft.com/azure/sentinel/best-practices


Final Architect Verdict

Analytics Tier is for action.
Sentinel Data Lake is for understanding.
Data Archive is for obligation.

Microsoft Sentinel is not a log storage platform — it is a security decision platform.

Correct tiering decisions determine whether Sentinel becomes:

  • A scalable SOC engine
  • Or an expensive log repository

This framework is designed to be used as a definitive Sentinel storage reference.


Leave a comment