Observations
It is no wonder that many organisations have struggled with bill shock regarding their Cloud-based VDI. This is certainly the case with ADS, which is relatively easy to set up, but can be difficult to accurately budget.
The costs associated when enterprises transition to the Cloud-based VDI are a complex concoction of:
- on-demand infrastructure charges, which are complicated by the inherent elasticity of such services, differences in compute, and both networking and storage cost structures.
- fees for licences associated with both VDI servers and clients, which may be a mix of concurrent usage, user-based, or even legacy perpetual licensing.
- endpoint services.
- often unknown operational and administrative costs.
- unexpected (if not unpredictable) demand on the service.
In addition, IBRS has noted that manage VDI service providers (both Cloud and on-prem) in Australia have frequently trimmed their costs of their proposals by reducing the capacity of the underlying infrastructure. This under-specification almost always results in a slow desktop environment and woeful user experience, that is then remediated with an upgrade to higher capacity servers, resulting in an associated increase in costs.
It is no wonder that organisations are increasingly seeking ways to rein in their Cloud-based VDI costs, and with ADS being top of mind, it is worth using this service as a guide to cost optimisation.
Best Practices in Microsoft Azure Cost Management
To start, it is necessary to establish a baseline of estimated costs for ADS, as well as a forecast of future needs. These estimated baselines should include consumption expenses such as costs in computing, shared services, domain controller, and user access, to name a few.
The Microsoft Azure calculator can help with this baseline calculation. However, any tool that helps organisations plan their VDI spending is only as good as the information being fed into it. Take care to correctly set the region where the organisational VDI infrastructure will reside and any geocentric replication needed. Determine the number of users, likely hours of VDI usage, and use this to experiment with different storage and bandwidth capabilities and review the budgetary impact. Finally, add allowances for other costs and errors in your assumptions.
Armed with the baseline, consider the following cost-optimisation tactics.
Decommission Overlapping Services with Other Vendors
While not directly related to VDI, significant savings can be made from decommissioning overlapping third-party services in favour of Microsoft services.
Many organisations are procuring Microsoft 365 licences, while not taking full advantage of the services provided and retaining overlapping third-party tools, the functionality of which could easily be assumed by the Microsoft service. This means they are effectively paying double for twice the trouble.
However, in some cases, it may be more practical to retain third-party products that overlap with services offered by an enterprise’s Microsoft E5 licence, rather than having to decommission the overlapping solutions.
IBRS emphasises the significance of preparing a critical plan to decommission such overlapping software, as these represent not only a licensing cost, but an operational cost as well.
Evaluate Current and Future IT Personnel Costs
Enterprises often commit the mistake of only considering the salary of in-house IT personnel who deploy, implement, and maintain Microsoft Azure. However, in regard to VDI, these costs should be considered against the comparable costs of either an on-prem VDI team (including both hardware infrastructure and software experts) and the desktop/end-user computing support team members.
Review Changes in Licensing Costs and Other Associated Fees
In the advisory article Microsoft 365: How Microsoft’s Evolving Strategy Impacts Your Future, IBRS observed that the company’s business model is anchored on leveraging its 365 subscription licensing that comes with bundling to develop a dominant tech stack from the users’ device into the Cloud.
The growing complexity of licensing on VDI has influenced many organisations to view the feasibility of the ratio of licences to seats with scepticism, including the value they receive from running concurrent sessions.
Microsoft’s clients must be informed about the company’s recurring licensing model that locks them in (which Microsoft has been implementing for decades already), as well as the high potential for licensing costs to hike over time. In addition, Microsoft has approved changes in its licensing conditions in 2019 that took out the provision to run virtual machines on rival hyperscale Cloud services such as Amazon Web Services (AWS) and Google, only to then offer licensing options for Azure. Therefore, organisations looking to migrate or extend existing on-premise VDI into the Cloud will need to carefully consider the licensing ramifications.
Use Automation to Cut Unused Capacity
Organisations do not operate at peak levels 24/7. In most cases, most people only use their computers for 10 hours a day, so there is no sensible reason to maintain a larger server for VDI support, whereas a smaller capacity will suffice.
Employing automation within Azure to reduce (and scale up) the specifications of VDI services can be highly effective in reducing costs. It has been reported that this simple tactic can see a 17–30 per cent saving when done well.
However, organisations should be careful to avoid allowing automation to deliver a VDI infrastructure that is under resourced, which will negatively impact user experience and their ability to work. There, it is important to also capture performance metrics and use this to inform the automated scaling processes.
Unfortunately, determining exactly what aspects of the infrastructure can be automatically scaled can be a challenging.
Tap the Cost Management Features from Third-Party Vendors
A number of specialised tools have emerged in recent times that not only support the automatic deployment of a new Microsoft Azure Virtual Desktop or Windows 365 environment, but also offer cost-reduction features. For example, Nerdio.
Compute-Based Strategies:
- VM Power Management: ensures the right virtual machine (VM) is turned on at the right time before the user needs it, and turned off when not in use.
- Just-in-Time Provisioning: refreshes all the VMs to their pristine state on a daily basis, so the users receive perfectly-configured, image-based, clean VMs without any configuration deviations from each other.
- 3-Year Reserved Instance Analytics: reserves excess capacity by analysing the autoscale history over the last 30 or 60 days, and makes the necessary recommendations based on potential savings.
Storage-Based Strategies:
- OS Disks Auto-Scaling: automatically changes the OS disk type to a lower SKU once VMs take a halt so that unused SSD storage is not paid for. This also does not impact user experience.
- OS Disks Shrink to 64GB: automatically downsizes disk space from the default 128 GB OS disk to 64 GB to save on OS disk storage.
- FSLogix Shrinks by 50 Per Cent: automatically adjusts provisioned quota on Azure Files premium shares to reduce storage.
In another interview with IBRS, Microsoft Technical Lead Pratima Singh stated, “Nerdio is mostly a management partner, I would say. So they are mostly into managing and optimising your deployment. They have deep expertise in Azure Virtual Desktop. They have done a lot of deployments across Australia. So we have partners who are certified on Azure Virtual Desktops.”
Next Steps
- Microsoft’s reference architecture for analytics caters for large organisations with an end-to-end approach. In many cases, it is overkill for mid-sized Australian organisations. Instead, develop a more limited approach, focusing on initial business needs and end-user adoption.
- Use cost comparison (monthly) calculators and document each significant iteration of options (in a saved spreadsheet).
- Consider the right storage disk for business needs: standard HDD costs are based on consumed storage. On the other hand, premium and standard SSD costs are based on allocated storage.
