Table of Contents
Purpose: Shows performance of all volumes belonging to a specific application. The scenario describes how to start a performance analysis from an identified spike in response time in the front-end
Pre-requisite: the Application groups need to be setup via the BVQ GUI first
There are two set of dashboards that support this analysis.
- Flip Application - customized to display performance based on group of applications
- VDisk by Application Analysis - performance details to help analyze a single VDisk belonging to an Application
You will need two Dashboard .json files. A license is necessary and can be downloaded here: Analysis Dashboards json file
1. Flip Application
2. VDisk by Application Analysis
Start the analysis by following these steps:
1.Import the two dashboard .json files
Make sure you are on the “Home” dashboard. Click on “Home”
A screen will open -on the right you will see “Import Dashboard”.
Click to upload file
1st select file “Flip Application” and make sure to select BVQServer in the “BVQ Server” field
Do the same to import 2nd dashboard file.
2.Open “Flip Application” Dashboard
Go to Main Dashboard Menu and select Flip Application
3.Select Cluster or Application to Analyze
Select a cluster and an Application, here
Once a specific application is selected, data will be loaded
- The dashboard displays general overview of application performance
For each application, the following performance information is displayed:
- Front-end IO/s performance gauge (left) - 1 hour mean value
2. Front-end performance graph of IO/s vs. latency
3. Front-end latency gauge (right) - 1 hour mean value
4. Data rate (blue) and transfer latency (white) by host
the thresholds have been set for transfer latency
5. Node ports load for all node ports which can be reached by any VDisks of the host (send and receive) and BCW% (yellow)
Thresholds are set for BCW%
Identify spikes in the front-end and follow the data path to check if this spike is repeated in other areas during the same time interval.
A good tip is to follow the pattern of the line. Check if it has the same pattern as the one line in the high peak time interval identified in the front-end. As shown in figure above.
Once a spike is identified, then do further analysis to identify single of group of VDisks involved by using the VDisk by Application Analysis Dashboard.
4.Start the Analysis for a Group or Individual VDisks belonging to Application
You can choose to then analyze a VDisk/ group of VDisks belonging to this Application.
To identify the single or group of VDisks involved in the high measurement spike in latency and validate the area where the performance issue is coming from.
For this, use the VDisk by Application Analysis Dashboard.
To open the "VDisk by Application Analysis", choose the button with the same name from the top right. Select this button to load the analysis dashboard.
This will load the dashboard in the same tab. It will not open in a different tab.
Follow these steps:
Step 1:Validate the Application you wish to analyze is selected in the Application field
Step 2: Select the time frame that corresponds to the time frame chosen in the "Flip Application" dashboard
Time period is selected from the top right corner
Step 3: Start with the first view “All VDisks from Application” view to get an overview of the different VDisks that are involved in the high measurement point of high latency. Hover over red line to determine which VDisk is creating a high peak.
Step 4: Select the specific identified VDisk from the VDisk Field
- From the VDisk Field, select a VDisk for deeper analysis
5. Detail Analysis of Individual VDisk
The Detail Analysis section consists of two sections: analysis and Data Rate and Cache
Open side tabs
1st section "analysis" contains 5 views, from top to bottom, left to right:
1. VDisk IO and Latency Use the mouse to see details
2. Transfer latency from storage to host
3. Data rate read and data rate write use mouse to see details
4. Host ports of all possible ports used by the VDisks data rate send and receive and latency plus BCW%. Use mouse to see details
5. MDisks used by the VDisks. Shows MDisk IO and latency
- Second section: Data Rate and Cache Section.
Consists of two views:
- Cache Hit and Cache Efficiency
White line represents Cache hit %
Blue filled area shows the incoming data into the cache and the outgoing data rate to the MDisks
Expand the window to full screen to see more details. The outgoing data rate should always be smaller than the incoming.
2. Cache partition of all MDisk Groups use min and max value (there at two MDisk Groups in this example)
Exception when the system performs many pre-stages
Result: From the detail analysis of VDisk, it was validated that the performance issue is related to a high Buffer Credit wait % which indicates there is a blockage in the SAN side.