Storage Performance

Troubleshooting Storage Latency with esxtop: The Admin’s Guide

Posted on April 11, 2015 Updated on April 15, 2026

When “the server is slow,” the storage subsystem is usually the first suspect. While vCenter performance charts are great for history, esxtop gives you real-time data from the heart of the hypervisor.

🛠️ How to Configure esxtop for Storage Monitoring

You can monitor performance at three different levels depending on where you suspect the issue lies.

1. Per-HBA (Host Bus Adapter) Mode

Command: Type esxtop, then press d.
Tip: Press Shift + L and enter 36 to see the full device names.
Fields: Press f and ensure b, c, d, e, h, and j are selected.

2. Per-LUN (Device) Mode

Command: Type esxtop, then press u.
Why use this? To see if a specific volume on your SAN is being hammered.

3. Per-VM (Virtual Machine) Mode

Command: Type esxtop, then press v.
Why use this? To identify the “noisy neighbor”—the specific VM that is consuming all the IOPS.

🔍 Analyzing the “Big Three” Latency Columns

To understand storage health, you must look at these three columns. They tell you exactly where the delay is happening.

Column	Name	What it represents	Threshold
DAVG	Device Latency	Time spent at the hardware level (HBA + SAN).	< 10ms
KAVG	Kernel Latency	Time spent inside the VMware VMkernel.	< 1ms
GAVG	Guest Latency	Total latency perceived by the Guest OS (DAVG + KAVG).	< 10ms

What the numbers are telling you:

High DAVG: The problem is external to ESXi. Check your SAN controllers, disk spindles, or fabric switches.
High KAVG: The problem is inside the host. This usually means the host is overloaded or there is a queueing issue (e.g., Disk.SchedNumReqOutstanding is too low).
High GAVG: Your users are feeling the pain. If this exceeds 10–15ms consistently, application performance will suffer.