Since its inception, Liquidware Labs has been heavily engaged in the transformation of the desktop. With visibility based upon leveraging our Stratusphere FIT product to support assessments, and Stratusphere UX to validate performance, we have a long history that encompasses over 400 metrics-driven desktop virtualisation projects. Agnostic of approach, the results of this history are unambiguous –success in VDI requires a new approach to how server-class resources are sized and implemented to support desktop workloads. Whether you invest upfront with an assessment or jump in and optimise later, there are critical attributes that must be considered.
Many early adopters of VDI assume that measuring desktop workloads prior to beginning a virtualisation project serves only to support the sizing and build-out of the host server environment. While this is partially true, there are other important benefits associated with this step, specifically as it relates to creating the optimal virtual machine (VM) image. We believe there are a couple of primary benefits to a metrics-driven approach to VDI, namely:
Capture the baseline user experience in the current environment – this is critical for ensuring that you provide an equal-to-better user experience when physical desktops are converted to virtual. This step also provides the ability to perform a before-and-after comparison of resource use and ultimately, end-user experience.
Monitor application use as it relates to desktop pools and images – this attribute of the assessment helps you to better understand what applications are used versus installed. This benefit also provides visibility into user and group resource requirements.
Regardless of when you measure, do not miss the critically important step of quantifying end-user computing resource requirements to support and ensure an optimal end-user experience. It is the cornerstone of a successful and optimised desktop virtualisation implementation. This paper will not detail the specific steps or process with respect to assessing user and machine-specific workloads. Rather, this paper will highlight the importance of memory and storage resources; specifically the critical role each plays in the end-user experience and overall performance of virtual desktop workloads.
Delivering end-user computing resources in a virtual desktop architecture is profoundly different from how we have provisioned, managed and optimised desktops in the past. More to the point, the way we manage server and storage resources to support VDI is about identifying and minimising resource bottlenecks in your environment.
We have found that the most common resource bottlenecks observed while leveraging the Stratusphere UX product relate to consolidation ratios, memory and storage resources. These are very common performance occurrences, which can be prevented during the early phases of implementation.
Poor consolidation ratio – this very common bottleneck is caused by unbalanced resource usage in host servers. Understanding how CPU and memory play a role in optimising VDI performance is critical to meeting total cost of ownership (TCO) and return on investment (ROI) goals.
Improperly sized VM memory – memory-to-disk swapping on the guest OS is another common issue that can cause end user performance issues in a VDI architecture. This can be especially tricky as host memory page sharing and ballooning do not prevent swapping if the guest VM OS “thinks” it is near capacity.
Storage and “boot storms”– successful VDI deployments also minimise the number of VM images required to satisfy all use cases. This desire can have the negative consequence of creating boot, or login storms, especially if storage requirements are not measured and allocated for both average and peak requirements.