One of the most difficult things to do in any
IT infrastructure is to establish what normal operations look like. The
extremes are always easy to spot, of course, but the heavy demand that
equates to normal business use is much more difficult to understand
without good visibility. Establishing normal operations can be more
complex in a virtual infrastructure but no less important. If you are
not sure what normal is, how can you properly identify what is
problematic and needs to be investigated before it impacts your end
users? This issue is critical in a virtual desktop environment because
often much higher expectations are placed on performance. I had a
conversation about this topic with a customer who had a mixture of
older and newer physical desktops, and what we concluded is that users
often do not report problems with physical desktops that are fixed with
a reboot; they simply shut it off and on, and if the problem goes away,
they consider it “fixed.” Ask yourself how often the computer logs of a
desktop are actually inspected for problems. When was the last time a
system administrator looked at the application or system log of a
Windows desktop?
With virtual desktops, however, these hiccups
get a lot of visibility in the early stages of adoption, so you need
high visibility on how everything is performing. This visibility must
include the View desktops, management components, performance of the
PCoIP protocol on the LAN and WAN, and the hosts and storage of the
vSphere environment. Monitoring all these pieces with traditional
monitoring tools can quickly produce a lot of information that becomes
difficult to correlate and understand. Traditional monitoring tools
based on static thresholds generate alerts on utilization or monitor
the Windows Event Viewer logs but do not provide a complete picture of
the how the entire environment is performing.
Although a number of
products from a variety of vendors monitor various aspects of the
virtual infrastructure, VMware has developed vCenter Operations to
bring this capability to VMware vSphere environments. It also has
developed custom dashboards for monitoring the performance of a VMware
View environment to extend the benefits to your virtual desktops. What
is unique about vCenter Operations (vCOPs) is that you get near
real-time, capacity, and historical performance and problem-tracking
information. The visibility extends to all components of your virtual
infrastructure, down to the storage and its dependencies. Many third
parties have also used the VMware APIs to extend monitoring and
alerting to their products, but vCenter Operations is tightly
integrated with VMware View.
vCenter Operations Manager aggregates
information that is collected by vCenter. For vCOPs, you need to ensure
that you are running VMware vCenter Server 4.0 Update 2 or later at a
minimum to be able to view storage system performance. Storage
visibility closes a huge operational gap in running large virtual
desktop environments. Often you see symptoms of underperforming
storage, such as delayed SCSI transactions in the Windows Event Viewer.
However, it is difficult to identify what the cause is. Because many
storage systems incorporate storage virtualization and distribute all
writes across all spindles, it is difficult to pinpoint the source of
the problem. It is rare now that a SAN administrator dedicates certain
drives to certain LUNs. This can make identifying storage bottlenecks
extremely difficult because SAN performance reports average performance
over time and across a large portion of the storage system.
vCenter Operations allows you to view
specific performance characteristics of your datastores in relation to
each other. VMware has extended vCenter Operations to aggregate
information from both the View Connection Server and Composer to
provide even greater visibility into your virtualization environment.
The benefits of vCenter Operations are many,
but rather than go through them all, let’s look at each in relationship
to the VMware View environment to understand how they deliver value.
The first thing you must do is install vCenter Operations and configure
the adapter to turn on the performance and monitoring of the VMware
View environment.
Installing vCenter Operations
vCenter Operations (vCOPs) Manager ships as a
virtual appliance and comes in several different versions; Standard,
Advanced, Enterprise, and Enterprise Plus. The fundamental difference
between each version is that Standard does not integrate the historical
trending, reporting, and capacity planning features. Advanced does but
does not allow you to customize the look, or dashboard as it is called.
Enterprise does allow you to customize the dashboard but is restricted
to virtualization environments only. Enterprise Plus allows you to
apply the tool to both physical and virtual environments. For a complete list of differences, refer to the VMware comparison matrix located at http://www.vmware.com/products/datacenter-virtualization/vcenter-operations-management/compare-editions.html.
vCOPs can be extended to input additional
metrics through the use of adapters. The visibility into VMware View is
facilitated through an adapter and the creation of seven VMware View
specific dashboards to display information such as user login times,
virtual desktop performance, and overall health of the environment.
Because it does install several custom dashboards, you do need
Enterprise to turn up these features. For an overview of vCOPS
integration, refer to http://www.vmware.com/products/desktop_virtualization/vcenter-operations-manager-view/overview.html.
Historically, the product was developed from
both Capacity IQ and vCenter Operations. In vCOPs 5, the two
independent products have become one. They are still two distinct
virtual appliances—the analytics and user interface (UI) appliances—but
they are deployed as a single unit or vApp. A vApp is a
collection of virtual machines that are grouped to allow common
configuration and to provide a complete service. For additional details
on vApps, refer to the vSphere 5 documentation available online at http://pubs.vmware.com/vsphere-50/index.jsp.
vApps were introduced as a method of
controlling multitier applications. In this case, the vApp allows you
to control the network, shares, and boot order as a single entity. In
the case of vCenter Operations, it ensures that the Analytics virtual
appliance is booted before the UI appliance.
vApps are designed to work with IP pools. IP
pools add a DHCP-like service that is associated with certain switch
ports. As with DHCP, you configure a range, gateway, and DNS settings.
After these settings are configured, you can tell the vApp appliances
to use them by editing its properties and changing the IP allocation
policy from Fixed to Transient.
The three types of IP allocation policies are
Fixed, which is a static IP assignment; DHCP, which tells the vApp that
a traditional DHCP service exists on the network; and Transient, which
tells the vApp that an IP pool is available to allocate network
information (see Figure 1).
Figure 1. IP allocation pool.
