Jul 24, 2015

EMC - VNX2 - Unisphere Analyzer

Unisphere Analyzer
Unisphere Analyzer is a block storage based performance tool used to monitor storage system performance. This performance tool has the ability to monitor the performance of array based LUNs, the Storage Processors (SPs) that own them, and their disk modules. Unisphere Analyzer gathers storage system performance statistics and presents them in various types of charts based on the choice of the user. Analyzer also has the ability to display performance data in real-time, or collect and archive performance data to be downloaded later. The information Analyzer provides can help you find and anticipate bottlenecks in the disk storage component of a computer system.
Unisphere VNX Client
EMC Unisphere VNX Client (commonly referred to as Unisphere off-array) is the locally installed, host-based version of Unisphere. Traditionally, a user would open a browser and connect to the IP address of the storage system. After connecting, the Unisphere interface would download to the user's PC over the network to enable management of the storage system. Unisphere VNX Client eliminates the download step as all Unisphere management components are already installed on the host.
Unisphere Server
In this Lab, you will use Unisphere VNX Client in conjunction with Unisphere Server. Unisphere Server is a host-based management server which has the ability to be configured as a domain master. This means the user has the ability to create a domain of storage systems under the host management server or by using the server as a stand-alone management server. In this Lab, you will log locally into the Unisphere Server, via the host's loopback address, and use Analyzer functions present in the software. Live Analyzer functions are unavailable when connecting to the loopback address, but the user can open previously collected Analyzer Archive files (.nar files).







Analyzer Settings
Section 1A focuses on the Settings section of the Analyzer window, which is located on the top left of the Analyzer
screen. In the Settings portion of the window, you will see options for:

  • • Performance Data Logging
  • • Customize Charts

Data Logging
Open the Performance Data Logging dialog box:
1. Click Performance Data Logging located under Settings in the Analyzer interface
A Data Logging dialog box will open. Because this is not a live system, the screen you receive is shaded like the one below. In the Data Logging window,there is a drop-down box to choose the Target array and customize options for Unisphere Analyzer. The options include:

• Real Time Interval
• Archive Interval
• Periodic Archiving
• Stop Automatically After
• Start
• Stop




Launch Help for Data Logging
As you are new to Unisphere Analyzer, the options in the Data Logging window may not be clear as to what they control. Complete the following to find out more information on the options in the Data Logging window:
1. Click Help



Customize Charts
Next, you will open the Customize Charts dialog box:

  • 1. Click Customize Charts within the Settings portion of the Analyzer window
  • 2. Check the Advanced option if not already checked
  • 3. Click Help


Customize Charts is the dialog box to set preferences for Unisphere Analyzer. The main tab in this window is the General tab, which can be seen when the window is launched.
The most commonly changed settings on the General tab are:

  • • The Advanced option within the Characteristics (Analyzer) box. The Advanced checkbox, when checked, will allow you to view Advanced Analyzer Statistics in the Performance Detail window. Without the box checked, you will just see basic statistics in the Performance Detail window.
  • • The Only get performance data for those objects that are selected in tree chart option within the Real-time Chart (Analyzer) box

For more information on each of the settings within the General tab, click Help. Close Help once you are done viewing it.



Survey Charts
1. Choose the Survey Charts tab
Survey Charts - Threshold options
The Survey Charts tab is the tab that is used in conjunction with the Performance Survey window. In the Threshold box, you can enable a statistic to monitor and specify the threshold value and the number of samples at that threshold. If the threshold is exceeded, the box for that characteristic in the Performance Survey window will be outlined in red. If the threshold is not exceeded, the outline will be green.
The options that are most often changed in this tab are contained in the Threshold box, and they are:

  • • Utilization [%]
  • • Throughput [I/O/sec]
  • • Bandwidth [MBs/s]
  • • Response Time [ms]
  • • Average Queue Length



Survey Charts - Graph Colors
The Survey Charts tab also contains a Graph Colors box. In this box you can change:
1. The Threshold Line color, which sets the color used to represent the threshold set when displaying data in the Performance Survey window
2. The Point controls the color used to for the data points in the Performance Survey window

Archives
1. Select the Archives tab
The Archives tab is where you would typically set the values once and not change them.
Within the Archives tab, in the General section, you can specify the Default Archive Location and the Default Dump Location. When opening Archives, setting the default path allows the user to quickly find their Archive files and open them.
1. Change the Default Archive Location to C:\Users\Administrator.VLAB\Desktop\Analyzer_NARs
2. Change the Default Dump Location to C:\Users\Administrator.VLAB\Desktop\Analyzer_NARs
3. Click on Apply
4. Click OK






Archive Management
Analyzer screen. In the Archive Management portion of the window, you will see options for:
• Open Archive
• View Archive
• Close Archive
• Retrieve Archive
• Merge Archive
• Dump Archive




Performance Sample






The new VNX systems provide an SP Cache Dirty Pages (MB) option as opposed to Cache % Full as in the previous generation of VNX. As you can see from the graph, there is an initial spike in the Cache Dirty Pages when the application started writing to MCC LUN 1, but soon after Multicore Cache started to throttle the workload and lower the amount of the dirty cache pages. Multicore Cache uses a different cache-cleaning model.
The system constantly evaluates the effectiveness of cache for specific workloads. Multicore Cache measures the rate of incoming IOs and monitors the underlying Pools and RAID Groups to predict its effectiveness for workloads. The second spike you see at the fourth time interval is where Multicore Cache is trying to see if it can take more writes from the host.The third spike that you see here is when IO was started to MCC LUN 2.




Review Disk Throughput - MCC Test Pool 1
Now let's take a look at the Disk Throughput for MCC Test Pool 1.
1. Click on a Disk
2. Select Total Throughput IO/s
MCC Test Pool 1 has 8 disks. Here, we can see the disks associated with the Storage Pool and the Total Throughput of these disks. As you can see, the throughput gradually increases until reaching a peak at around 440 IOPs. All the disks in the Storage Pool also have similar throughput, confirming the IO is distributed equally among the 8 disks.

No comments:

Post a Comment