Description:

During performance validation for MTV 2.10, significant degradation was observed in total migration duration when running large-scale migration scenarios involving multiple ESXi hosts and concurrent VMs.

The degradation is consistent across multiple cycles and appears to be correlated with increased DiskTransferV2v phase durations.
This behavior was not observed in MTV 2.9, where overall migration times were significantly shorter under identical test conditions.

Test Scenarios Impacted:

TC 2.8 – 80 VMs, 1 Disk (50GB, 35% usage), 8 ESXi Hosts

Avg Migration Duration:

• MTV 2.9 → 0:37:32

• MTV 2.10 → 0:45:33

TC 3.5 – 50 VMs, 1 Disk (50GB, 35% usage), 5 ESXi Hosts

Avg Migration Duration:

• MTV 2.9 → 0:21:32

• MTV 2.10 → 0:36:07

Key Observations:

• The DiskTransferV2v phase shows a ~40–50% increase in duration across both test cases.

• The degradation scales with the number of concurrent ESXi hosts and VMs, indicating possible contention or scheduling inefficiency introduced in MTV 2.10.
• No infrastructure or network-level changes were introduced between 2.9 and 2.10 cycles.

Environment Details:

• • vCenter: VC7-1

• • DataStores:

• PerfTest_VC7_1_FC_10TB

• PerfTest_VC7_1_ISCSI_24TB

• • ESXi Hosts: 5 - 8 hosts per test

• • Each VM: RHEL 7.9, 50GB disk (35% usage)

• • Migration type: Cold

• • Parallel VM limit: Max 10 per ESXi (20 in-flight total)

Suspected Area:

• • Possibly introduced regression in I/O scheduling or throttling logic between MTV 2.9 and MTV 2.10.
  • Potential bottleneck in DiskTransferV2v phase handling when distributing multiple concurrent VM transfers across multiple ESXi hosts.

• • It’s also suspected that additional internal validations, conditions, or checks were added in MTV 2.10 compared to 2.9, which could be increasing processing overhead during the migration flow especially under multi-host and high-concurrency conditions.