SLA Granularity Adequacy Index (SGAI)

GMI-TEL-031TelecommunicationsInfrastructure-Dependent

Definition

Percentage of application-experienced quality events (latency spikes, throughput drops, session interruptions) that are visible at the SLA monitoring system's measurement granularity, measuring whether the monitoring is fine-grained enough to detect what applications actually experience.

The Measurement Gap

Measurement Gap and Decision

Decision Supported: Is the SLA monitoring system seeing what the application sees, or is it averaging away quality events that matter?

Measurement Gap:

SLA monitoring systems in multi-slice environments typically measure performance at 1-minute to 5-minute averaging intervals. An application may experience a 500ms latency spike that the 5-minute average smooths to an imperceptible blip. For eMBB applications, this may not matter: the user barely notices. For URLLC applications (industrial control, remote surgery, autonomous vehicle coordination), a 500ms spike is a catastrophic event that the SLA monitoring system literally cannot see. The operator reports 100% SLA compliance while the application experienced multiple safety-critical quality failures. The monitoring system is certifying compliance at a granularity that cannot detect the failures that matter. SGAI quantifies this gap by deploying fine-grained monitoring alongside the standard SLA system and comparing detection rates: revealing exactly how much of the application's reality the SLA system misses.

Citations:

  • 3GPP TS 28.554, "5G End-to-End KPIs," Release 18: defines performance measurement intervals without assessing whether those intervals are adequate for the application types they monitor.
  • GSMA, "5G Assured: SLA Specification and Monitoring," 2024. Specifies SLA monitoring architecture without addressing the granularity adequacy problem for latency-sensitive applications.

Formula

Plain Language: Deploy fine-grained (sub-second) application-level monitoring on a sample of slices. Compare the number of quality events detected at application granularity to the number detected by the SLA monitoring system's standard granularity (typically 1–5 minute averages). SGAI is the ratio.

Notation:

SGAI = Events_detected_at_SLA_granularity / Events_detected_at_application_granularity × 100

Variable Definitions:

  • Events_detected_at_SLA_granularity = quality events (threshold crossings) detected by the standard SLA monitoring system in its standard measurement intervals (count)
  • Events_detected_at_application_granularity = quality events detected by sub-second application-level monitoring for the same workloads and time period (count)
  • A "quality event" = any period where the metric exceeds the SLA threshold, regardless of duration

Units: Percentage. 100% = SLA monitoring detects every event the application experiences. 50% = SLA monitoring misses half the events due to interval averaging.

Sign Convention:

Higher is better. 100% means the SLA monitoring system has adequate granularity.

Edge Cases:

  • Sub-second monitoring may detect noise that is not meaningful to the application: define a minimum event duration (e.g., greater than 100ms) to filter transient spikes.
  • Different applications have different sensitivity: compute SGAI per application class (URLLC vs. eMBB vs. mMTC).
  • Fine-grained monitoring is expensive. This is a sampling exercise, not a permanent deployment. Sample 5–10% of slices per quarter.
  • Minimum sample: 50 quality events detected at application granularity per measurement period.

Proposed formula, pending validation pilots.

At a Glance

Unit
Percentage
Trend Direction
Higher
Reporting Frequency
Quarterly
Category
SLA Assurance Integrity and Structural Sustainability
← SLA Assurance Effort Trend (SAET)Slice Instantiation Timing Accuracy (SITA) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment