Home > Tintri VMstore™ > Knowledge Base > Understanding the latency dashboard on the VMstore UI

Understanding the latency dashboard on the VMstore UI

Applies To

 

Product(s)

Version(s)

VMstore

Tintri OS 4.x

Description

You can access the latency dashboard on the VMstore UI by clicking on the latency tab on the main dashboard.

 

Note: The Flash hit ratio tab will only display on hybrid arrays. 

 

VMstore performance

 

The latency dashboard that opens up contains two sections, a graph on the left and a list of VMs on the right. 

Details

The graph on the left is a display of total system latency on the VMstore. This latency is calculated as the sum of all read and write operation latency in a 10 minute period, divided by the total number of operations within that 10 minute period. It is effectively a weighted average of the VM latencies by IOPS but it also includes non-VM read and write operations.

 

The latency measurement is broken down by the following latency values:

  • Host: The latency attributed to the host.  This value is calculated by subtracting the network and storage latency values from the “end-to-end” latency measurement.
  • Network: The latency attributed to the network as measured at the RPC layer.
  • Storage: The latency attributed to and measured within the Tintri file system, including disk latency and contention.

While viewing the Tintri latency graph, holding the mouse pointer over any data point will display the individual latency values. Note that in older releases, there was a separate Disk category that is now included in the Storage category. 

 

Some more details about the graph:

  • There is a grey vertical bar on this graph which indicates realtime position. 
  • The graph can be dragged left or right to review historical data. 
  • The blue vertical bar on the graph will cause the list of VMs on the right to display system latency contributions for that particular point in time.
  • The graph will overlay the name of the VM contributing the most to system latency at that point in time.

 

The list on the right is a list of VMs sorted by largest contributor to system latency. This is calculated as a product of IOPS and Latency, with the highest values at the top of the list. 

 

For example, looking at the VMs in the screenshot: 

VM

IOPS Latency Value
RD-Tintri Global Center3.5 101 0.4 40.4
VMware vCenter Server Appliance 32 0.5 16
VMware vRealize Log Insight 11 0.5 5.5

 

 

 

Viewing 2 of 2 comments: view all
Very nice Andrew. Great technical detail especially for the new folks
Posted 16:21, 18 Apr 2017
Final approval done
Posted 06:48, 25 Apr 2017
Viewing 2 of 2 comments: view all
You must to post a comment.
Last modified
12:45, 12 May 2017

Tags

Classifications

This page has no classifications.