Why Virtual Server Monitoring Tools are Not Sufficient for VDI Monitoring?
The use of virtualization for hosting server VMs is well understood. Today, most virtualization platforms provide administration and monitoring software that admins can use to track the health of their hypervisors and virtual machines (VMs).
What is VDI?
The use of virtualization for desktops is more recent. For desktop virtualization, a virtual desktop infrastructure (VDI) hosts a desktop operating system within a virtual machine (VM) on a centralized server and streams a pre-built desktop image to remote endpoints: desktops, laptops, thin clients, and mobile devices.
Users at the endpoints can access the virtual desktop, which has an operating system and applications that need to work as they would from their local workstation.
Eliminating the need to install expensive and resource-intensive applications and hardware at every user endpoint, VDI cuts down IT investments and simplifies accessibility for end-users.
As virtual desktops are implemented as virtual machines, many IT Operations teams try to use their existing virtual server monitoring tool (e.g., VMware vCenter, Citrix XenCenter, Microsoft SCVMM, etc.) to monitor a virtual desktop infrastructure. While virtual server monitoring tools can be used to monitor the VMs and hypervisors supporting VDI, they do not provide the level of visibility and coverage necessary for true end-to-end performance monitoring, diagnosis, and reporting for VDI.
Comparing the Monitoring Needs of Server Virtualization and VDI
The table below compares the monitoring needs of virtual server and virtual desktop infrastructures and highlights why a virtual server monitoring tool is not sufficient for monitoring virtual desktop infrastructures (irrespective of whether VMware Horizon or Citrix Virtual Desktops is used).
|Monitoring, alerting, and reporting is mostly from the VM perspective. For e.g., which VMs are on, what resources (CPU, memory, disk) are they using, etc.
|Monitoring must be from the user perspective. To handle support calls from users, it is imperative to know which users are logged in, which VM a specific user is assigned to, what applications he/she is accessing, and what resources the user is consuming. Even from a reporting perspective, it is important for reports to be user-based – for example, who are the top VDI users that are consuming resources?
|Key Components to be Monitored
|The virtualization platform or the hypervisor (e.g., VMware vSphere, Citrix XenServer, Microsoft Hyper-V) and the virtual machines (guest VMs) are the main components to be monitored.
|The virtualization platform and the virtual machines are important components of a virtual desktop infrastructure. But there are many other tiers that are almost as important – for example, connection broker, StoreFront, Active Directory, profile servers, provisioning servers, data stores, etc. The network connectivity from the user terminals to the server farm must also be monitored as it can affect user-perceived performance.
For virtual desktop success, users must be happy with the experience they are getting when accessing their desktops. Therefore, measuring the various aspects of end-user experience (e.g., logon times, application launch times, screen refresh latency, etc.) is very important.
|Virtual Machine Workload
|The workload of a VM is similar over time because the application servers deployed on a VM do not change over a few hours.
|The workload of a virtual desktop varies depending on the user who is logged on to the desktop. Depending on the user’s role in an organization, the application mix running in the virtual desktop could be very different. Therefore, monitoring should be based on user activity, and not on virtual machine activity.
|Number of Virtual Machines to Monitor
|There are only few VMs to monitor: <10 per physical server. Further, the VMs remain powered on all the time as the application servers they host need to be accessible always.
|There are many VMs to monitor: 30–70 per physical server. The VMs are powered on when users log on and then powered off when users log off. From a monitoring perspective, the monitoring solution must be scalable so that it can accommodate the larger number of VMs in operation.
|Monitoring Applications Inside Virtual Machines
|To obtain in-depth monitoring of applications (such as Citrix, Oracle, etc.) that are running on the VMs, it is likely that monitoring agents need to be installed in every VM.
|Virtual desktops do not have server applications running on them. So, in-depth monitoring of applications on the desktop is not required.
Furthermore, the greater density of VMs on each physical server means that the cost of installation, licensing, and maintenance of the agents is higher for VDI. The resource overhead resulting from agents on every desktop is also high.
Therefore, from a monitoring perspective, the challenge is to monitor activities inside a VM (e.g., application launches, resource usage of various applications, etc.) without relying on agents to be installed in the VMs.
Key Requirements for Monitoring VDI
For end-to-end monitoring of VDI, organizations need:
- Monitoring of all aspects of user experience including logon time, desktop boot time, application launch time, screen latency, etc. Synthetic and real user monitoring approaches can be used for this.
- Monitoring of every layer and every tier of VDI: A typical virtual desktop infrastructure includes many tiers of software and hardware. Load balancers/secure gateways (e.g., Citrix NetScaler, F5 BigIP, etc.), StoreFront/web frontend servers, the connection servers, profile management servers, license servers, streaming servers (e.g., PVS), SQL datastores, app layering systems, etc. all have to be monitored.Any abnormalities in these tiers can manifest as slow desktop experience for the end user. As the function of each tier is unique, the monitoring system needs to have customized monitoring for each tier.
- Monitoring of VDI as a service, not as a silo: When a user complains about slow desktop performance, he/she is referring to the VDI service – not just about the desktop. Abnormalities anywhere in the infrastructure – network, server, storage, virtual platform, etc. – can manifest as slow access to the end user. For administrators to be able to troubleshoot VDI problems quickly and reduce MTTR, they need end-to-end visibility. Color-coded topology views with autocorrelation of alerts can help them quickly identify where the real issues lie.
- Capacity optimization and planning: VDI administrators are often asked to determine how they can support additional users. They need access to ready reports that help determine how they can get more out of their current investments – e.g., by knowing which are the top resource consuming applications, how well is load balancing working, etc. At the same time, they need access to reports that support “what if” scenarios (e.g., what if 1000 more users have to be accommodated?). Capacity planning reports can provide VDI admins with the insights they need to make intelligent investment decisions.
VDI Monitoring with eG Enterprise
eG Enterprise is a purpose-built solution that addresses the monitoring needs of both desktop virtualization and server virtualization.
- With dedicated monitoring models for Citrix Virtual Desktops and VMware Horizon, eG Enterprise provides deep visibility to uncover complex performance issues in VDI environments. Learn more about eG Enterprise for VDI Monitoring »
- A patented inside-outside monitoring approach provides comprehensive VM, host, and hypervisor metrics for server virtualization environments (VMware, Microsoft, Citrix, Nutanix, Red Hat, and more). Learn more about eG Enterprise for Virtual Server Monitoring »
Flexible deployment options allow monitoring of VMs in server virtualization environments either with or without agents inside each VM. For desktop virtualization, without installing an agent in every VM, organizations can obtain the performance visibility and user experience insight that is needed for virtual desktop deployment success and end user satisfaction.