UCS Blade Utilization / Performance Monitoring Tools?

[flbog.irm]

An unnamed software system vendor from whom we have purchased product was not ready to support their product in a virtualized platform environment. But we are a Cisco UCS shop. So we added a chassis, put four blades in it, and let the vendor install and configure their product across the four blades. I am not satisfied with the performance of their application, and have doubts that it is actually using the resources of all four blades. And I have no idea how to see the utilization of CPU Mhz or memory or even the blade-to-blade communication that one would expect the software system's clustering to implement.

NetApp has a great tool in OnCommand System Manager for monitoring the storage system performance. vSphere does a nice job with vSphere Client for datacenter hosts and the guests defined therein. But these blades are used as individual Linux servers....I'm looking for a way to get a view of how the application is using these compute resources.

Is anyone aware of a useful performance dashboarding tool available for monitoring the performance of blades in a UCS chassis? 

[Walter Dey]

Cisco UCS Central 1.2.1a enables you to generate standard and customized reports from the Statistics tab. You can generate reports on the following data in the registered Cisco UCS domains:

- Cooling
- Network

- Power
- Temperature

UCS Central features two report options:
Standard Reports.Predefined reports on Peak Fan Speed, Received Traffic (Rx), Transmit Traffic (Tx), Average Power, and Peak Temperature. You can run any of these predefined reports any time. You can also modify the predefined configurations, but cannot create any new standard report.
Custom Reports. Option to create customized reports from any of the available report options. Based on your requirements, you can create either individual reports or groups of reports. You can create, edit, or delete the custom reports at any time.

There is a nice lab where you can test this: go to dcloud.cisco.com and login with your CCO account.

[Kent Erickson · Accepted solution]

Cisco UCS Performance Manager can track the resource utilization of Linux operating systems running on bare metal servers in a UCS domain.

UCSPM collects:

• CPU and memory utilization from the Linux OS
• LAN and SAN utilization (receive and transmit) for each blade
• vNIC and vHBA utilization for each service profile on each blade
• LAN and SAN utilization (receive and transmit) for each chassis
• Remaining fabric capacity for each chassis
• Fabric utilization for each IO module in each chassis

Between CPU, memory, and multi-tier fabric utilization, you should have a pretty good idea whether UCS resources are constraining the application. Seems pretty unlikely, but you never know!

I've attached a screen capture of a bare metal server, showing a bit of this. It's running Windows and the metrics will vary a bit for Linux.

[Walter Dey · Accepted solution]

Thanks Kent for the hint

Just for the community:

This is a brand new product, released October 2014.

It is CentOS based, runs as a VM under ESXi or Hyper-V, and requires a license.

I have some questions, eg.

- which ESXi versions are supported, 5.1, 5.5 U1, 5.5 U2,....

- which Window server versions: 2012, 2012 R2,....

- license per UCS domain, cost ?

- positioning against DCNM-SAN, DCNM-LAN

[Kent Erickson · Accepted solution]

A couple of answers:

• ESXi hosts are monitored using a connection to vSphere, not directly to each ESXi host. By connecting to vSphere, UCS PM is able to track VMs as they vMotion from host to host. vSphere versions supported are 4.1, 5.0, 5.1.
• Microsoft Servers are monitored using the WinRM protocol. Supported versions include 2003, 2008, 2008 R2, 2012, and 2012 R2. Hyper-V is supported on 2008, 2008 R2, 2012, and 2012 R2.

UCS PM is a performance and capacity monitoring product for people running UCS-based integrated infrastructures. That means it covers OS, virtualization, UCS compute, storage, and a focused set of network devices.

DCNM helps people implement, visualize, and manage a UCS Unified Fabric. It works with a really broad range of network devices.

You might well want both products. It depends on what you're trying to do!

In general UCS PM is licensed by the server, where all servers in a domain must be covered. There's one version just for the UCS fabric and one that covers all the integrated infrastructure components.

Perhaps a Cisco employee can respond on pricing.

[vijakka]

UCS Performance Manager is the new product in the UCS software family to address the health and performance monitoring related questions our customers have. Please visit

cisco.com/go/ucsperfmgr for more information on the product. There you will see product documentation - install and administrator guides with detailed information. We will be adding additional information as we go along.

As mentioned earlier UCS Performance Manager is licensed per UCS server in the UCS Domains being managed. The product PIDs are UCS-PM-IE and UCS-PM-EE, available to order.