Definition
Aggregate application-level disruption caused by workload migration events, measured in service-quality-seconds lost, compared to the infrastructure efficiency gained by the migration.
The Measurement Gap
Measurement Gap and Decision
Decision Supported: Are workload migrations worth the disruption they cause, and for which workload types should migration be restricted?
Measurement Gap:
MEC orchestrators trigger workload migrations to rebalance load across edge nodes. Moving workloads from overloaded nodes to underutilized ones. The orchestrator measures the efficiency gain: better utilization, lower aggregate latency, reduced hotspot risk. What it does not measure is the disruption experienced by the migrated workload during the migration itself: latency spikes during state transfer, dropped sessions during the cutover, state synchronization gaps that cause application errors. The efficiency gain is measured in the orchestrator's domain (infrastructure metrics); the disruption cost is experienced in the application's domain (user-facing quality). These two domains are never compared. An orchestrator can report "successful migration, utilization improved by 15%" while the migrated workload experienced 200 dropped sessions and a 3-second latency spike. MDCI bridges this gap by converting both the disruption cost and the efficiency gain into a common unit (service-quality-seconds) and computing the ratio: revealing whether the migration was net-positive or net-negative from the application's perspective.
Citations:
- ETSI GR MEC 035, "MEC in 5G Networks," 2024: describes workload migration mechanics but does not measure application-level disruption caused by migration events.
- IEEE, "Live Migration of Virtual Network Functions: Survey and Open Challenges," IEEE Communications Surveys & Tutorials, 2023. Catalogs migration techniques without defining a disruption-to-benefit ratio metric.
Formula
Plain Language: For each migration event, measure the application-level disruption (latency spike duration multiplied by magnitude, dropped sessions, state synchronization gaps). Convert to a common unit (service-quality-seconds lost). Compare to the efficiency gained (resource utilization improvement multiplied by duration of improved state). The ratio tells you whether the migration was net-positive or net-negative.
Notation:
MDCI_event = Disruption_SQS_lost / Efficiency_SQS_gained
MDCI_period = Median(MDCI_event) across all migration events
Variable Definitions:
- Disruption_SQS_lost = service-quality-seconds lost during migration, calculated as latency exceedance multiplied by duration plus dropped sessions multiplied by session value (SQS units)
- Efficiency_SQS_gained = resource utilization improvement multiplied by time the improvement persists before next rebalancing event (SQS units, normalized to same scale)
- SQS (service-quality-seconds) = common unit normalizing latency, throughput, and session continuity into a single disruption measure
- Median used for the period aggregate to avoid outlier sensitivity
Units: Dimensionless ratio. MDCI below 1.0 = migration was net-positive (efficiency gained exceeds disruption). MDCI above 1.0 = migration cost more than it saved.
Sign Convention:
Lower is better. MDCI of 0 represents perfect seamless migration with zero disruption.
Edge Cases:
- Migrations with zero measurable disruption (seamless live migration): MDCI = 0. Perfect migration. These should be the majority.
- URLLC workloads where even 50ms disruption has real-world consequences: apply a severity multiplier to the disruption calculation for URLLC class.
- Efficiency gain is ephemeral (load rebalances naturally within minutes): short-lived efficiency gains should be discounted in the calculation.
- Minimum sample: 30 migration events per measurement period.
Proposed formula, pending validation pilots.