Azure Windows Virtual Desktop (WVD) has changed the game for remote working. Affordable and easier to set up and manage than traditional solutions, it makes virtual desktops accessible for all businesses. So how do you go about optimising Azure WVD?
In our first two blog posts in this series we discussed why you would use Azure WVD and how to set it up. This third blog post gives the IT lead in your organisation high-level guidance on optimising the cost and performance of Azure WVD according to your particular business needs. The only IT knowledge you will need to manage this environment is traditional desktop management skills.
You may be looking for a virtual desktop or remote app solution or your organisation may have already set up Azure WVD, we outline how you can make the most of Microsoft’s best-in-class solution.
Want advice from a trusted Azure WVD expert? We are an Azure Gold Partner and can help. Book a free, no-obligation consultation or read on to find out about the key considerations for optimising Azure WVD in your business.Azure WVD - Book your free consultation
As many IT professionals will know, there is often a trade-off between cost and performance – if you want better IT performance, it usually costs more. While this is true for many elements of Azure WVD, there are some key ways to optimise the performance of your Azure WVD environment that won’t cost a penny.
For optimizing Azure WVD, there are two, key performance considerations – latency and speed.
Latency is the round-trip time (RTT) and primarily based on the physical distance from your users to the virtual desktop session host (or server) and back again. All things being equal, the latency decreases the closer the network points of a connection are to each other. Therefore, it is important to consider which Azure region your session host is deployed in.
For a decent virtual desktop experience, the round-trip latency should be less than 100ms. Between 100ms and 200ms is still reasonable, however anything above 200ms will be noticeable by the end-user.
The Microsoft WVD Experience estimator allows you to estimate the connection round-trip time (RTT) from your current location, through the Windows Virtual Desktop service, to each Azure region in which you can deploy virtual machines.
For example, if your users are in the UK, logically you’ll choose a UK data centre for your session hosts. If you also have a set of users on another continent, then you can deploy their session hosts in an Azure datacentre that’s local to them.
However, if you have a set of users on another continent that need access to back-end systems in the UK, you’ll need to balance the user latency to the virtual desktop and the virtual desktop latency to the back-end system. Therefore, you may want to consider locating your session hosts between the back-end system and your users. Our Azure WVD specialists can help you identify the right option for your requirements.
RDP Shortpath is a new Microsoft service in public preview designed to establish direct connectivity between the remote desktop client and the session host with a number of potential benefits:
The removal of additional relay reduces RTT, which can improve user experience with latency-sensitive applications and input methods.
The second key consideration is the speed at which your virtual desktop is running your applications and performing tasks such as opening Outlook or creating a new Word document for example.
The speed your virtual desktop runs at is determined by how resource intensive the workload is (according to the applications that you’re running and how intensively you use them) and how much compute resource (processors and memory) you allocate to each user. The more compute resource you allocate to a user, the faster their virtual desktop will run.
To help estimate the amount of compute resource you need, Microsoft have published this useful guidance on the types of remote desktop workloads from ‘light’ through to ‘power’ with corresponding recommendations for virtual machine sizing.
For example, if your workloads are deemed to be ‘heavy’ then the recommendation is a maximum of two virtual desktop sessions (users) per vCPU, with a minimum virtual machine specification of four vCPUs and 16 GB of RAM such as the D4s_v3, which is a commonly used virtual-machine size for WVD session hosts. Therefore, you could have eight virtual desktop sessions on this session host. Our Azure WVD consultants will advise you on how best to scale your deployment depending on the expected need of each type of user.
Once you’ve deployed your virtual machines, the first priority is to test their performance for your users. If the virtual desktop sessions are performing well, you can try increasing user density per processor – either by allocating more users per session host or by decreasing the size of the session host. Whilst this decreases your cost, you will need to monitor and ensure the trade off on performance.
Windows 10 comes with a number of preinstalled applications. Removing those that aren’t wanted and not running unnecessary processes will make virtual desktop sessions more efficient. This will improve performance for the allocated compute resource, or alternatively, you can reduce the amount of compute resource and its costs accordingly. By optimising in this way, you can increase user density by up to 15%. Watch this 10-minute video from the Azure Academy for a useful overview of this optimisation plus Virtual Desktop Optimization tool with the scripts and instructions you’ll need.
NB You can’t use this code to optimise the image itself as some of the code won’t survive the sysprep process, instead run the code against the Azure WVD session hosts once you’ve deployed them from the image.
If you’d like to find out more about the most suitable options for your business to ensure an efficient network connection, our Azure WVD consultants will be happy to help.Azure WVD - Book your free consultation
Your Azure virtual machines or session hosts represent the primary cost of an Azure WVD environment. Therefore, increasing user density as discussed above will reduce your costs. Once you’ve identified your optimum user density, there are steps you can take to reduce your costs further.
Azure virtual machines are billed on a pay-as-you-go basis, so there can be considerable cost savings if you’re able to switch your machines off outside working hours. You’ll continue to pay for any virtual machine discs, however this is typically a minor cost in comparison to the virtual machine.
If you only need your Azure WVD environment to be available to users during office hours i.e. 08.30 to 17.30, Monday to Friday, and you’re able to switch off your virtual machines outside those hours, then you may pay as little as 27% of the full monthly virtual machine cost.
Microsoft provides an autoscaling solution for Azure WVD. This automatically turns session hosts on and off and manages user sessions for you. Whilst there are some limitations, autoscaling allows you to establish core hours so that you only pay for what you use:
Read this Microsoft article for a detailed explanation of how to scale session hosts using Azure Automation or for a refresher on the some of some of the typical technology and concepts involved, read our article How to implement Azure Windows Virtual Desktop (WVD) | Compete366.
By default, all the resources you deploy in Azure are Pay As You Go (PAYG) and without minimum commitments or tie-ins. However, Azure also offers other pricing options such as Reserved Instances (RI) for virtual machines. This means that you can benefit from a significant discount in return for committing to resources for one or three years. You can still pay monthly and have the flexibility to cancel within the agreed period, subject to a cancellation fee.
The key point to note is that if you buy an RI for a virtual machine, then you pay for it whether you turn the virtual machine on or not. So the idea is that you apply RIs to the virtual machines that are going to be on all the time, as you don’t make any saving by turning them off with RI pricing.
For further details of how to save more with RI read Microsoft’s article Azure Reserved Virtual Machine Instances | Microsoft Azure.
We normally recommend:
There is also a crossover between the two pricing options that you may need to consider. If you get a 40% discount for a one-year RI on a particular virtual machine, then that machine would need to be switched on for more than 60% of the time for RI to be cheaper than PAYG. For example, if the PAYG price of the virtual machine is £100 per month and it’s on for 50% of the time, then the cost would be £50. If you applied the RI pricing option, then the cost would be £60.
You can mix and match PAYG and RI pricing, so for example if you always wanted to leave two session hosts on and turn the rest on and off, then you could use RI pricing for two and PAYG for the rest.
If you need to weigh up how best to optimise Azure WVD costs and performance for your business, then please contact us for a free discussion.
As Azure Gold partners, our certified consultants will work with you to identify and implement the right solution for your business, with guidance on all Azure and Microsoft Office 365 elements. It means that the only IT knowledge you’ll need to implement and manage this environment are traditional desktop management skills.