Avoiding Bill Shock
So, you’ve made the decision to utilise virtual desktops in the cloud, specifically AWS WorkSpaces. How do you make sure those forecast cost benefits are realised?
People often misinterpret the “pay for what you consume” message when cloud computing is discussed. This is true to a point, largely where serverless computing such as AWS Lambda is concerned, but for pretty much everything else you also pay for anything you provision.
With this in mind, here are some of the key areas to consider:
Size WorkSpaces Correctly
The priority here is knowing your users, more specifically their task profiles. The needs of a worker consuming predominantly web-based applications can be met with a lower specification bundle, whereas someone working with large spreadsheets and reporting tools will need much more computing power.
There are 7 Windows bundles and 5 Linux bundles, ranging from 1 vCPU/2GB RAM all the way to 16 vCPU/122GB RAM/1 GPU/8 GB video RAM, so there is plenty of flexibility.
Choose the Most Appropriate Billing Option
This is more of a business profile-driven decision, and again there may not be one correct choice.
A Standard (2 vCPU/4GB RAM, 80GB root volume/10GB user volume) Windows WorkSpace bundle running in Ireland costs:
Always on: $34/month
Hourly: $8/month + $0.32/hour (the monthly charge covers the provisioned storage)
As an extreme example, a WorkSpace provisioned for hourly billing that is used constantly for an entire month would cost around $240 – more than 7x the always on cost. Remember you are in control of the billing model choice; AWS simply produces the bill based on this!
So, what is the cut-off point where always-on billing is the more cost-effective option? Let’s do the maths with the same example WorkSpace bundle:
Subtract the standing monthly charge from the always on cost: $34 – $8 = $26
Divide this number by the hourly on-demand cost: $26 / 0.32 = 81.25
So, if an on-demand WorkSpace is used for more than 81.25 hours in a month then Always On is a more cost-effective billing model.
AWS have produced the WorkSpaces Cost Optimisation tool as a quick-start template. This collates metrics from your WorkSpaces environment and makes billing model recommendations based on the previous month’s usage. You can even let it make billing model changes to individual WorkSpaces if you wish.
Unfortunately though, with the user and business profile considerations highlighted above there is almost certainly additional work to do with the tool’s output in order to ensure the recommendations are right for your business.
Windows or Linux?
It might be obvious but running a Windows-based AWS WorkSpace is more expensive than the Linux equivalent. This is because the required Microsoft license is factored into the WorkSpace running cost.
Whilst general office workers will likely need a Windows operating system for familiarity, application compatibility, etc. it’s very likely your development teams will be much more comfortable with a Linux-based environment.
You should also keep in mind that whilst the Windows AWS WorkSpace has a Windows 10 look and feel, it is actually Windows Server 2016 under the covers. If you want a true Windows 10 experience for your WorkSpaces, then you have to run a minimum of 200 WorkSpaces on dedicated AWS hardware and bring your own licenses (more details here: https://docs.aws.amazon.com/workspaces/latest/adminguide/byol-windows-images.html). This does save $4 per month per Windows WorkSpace, however there are additional operational overheads such as image and license management which have to be factored in.
Where to Provision WorkSpaces
The cost of foundational resources for WorkSpaces (compute, storage, power, etc.) vary from region to region, so this is definitely something to consider.
AWS have a tool available (https://clients.amazonworkspaces.com/Health.html) that recommends the best AWS region to run your WorkSpaces in based on network speed and latency from where the tool is run.
This tool illustrates that the London region is the fastest and therefore recommended region, however if you are able to use one of the alternate regions in Europe then costs are lower. Running a standard Windows WorkSpace bundle (2 vCPU/4GB RAM, 80GB root volume, 10GB user volume) in an always on configuration costs as follows:
So, for a 100-seat deployment you could save $9,600/year simply by running in Ireland rather than London.
Manage Leavers Appropriately
Having a solid organisational process in place to manage leavers is an often-overlooked aspect of managing down your AWS WorkSpaces spend. Laptops and mobile phones are tangible, physical assets that can be redeployed to new users, whereas a WorkSpace is associated to exactly one user.
A forgotten always-on, Standard Windows WorkSpace running in Ireland will incur a $408/year AWS cost.
Regularly Review your Spend and Utilisation
Taking advantage of the inherent flexibility from cloud computing is equally important with AWS WorkSpaces. If your finance team need to run some abnormally large and intensive end of year processes, simply increase their WorkSpace specification for the reporting period and reduce it again afterwards.
Likewise, the Cost Optimiser output should be regularly reviewed to ensure the selected billing model for each WorkSpace is still optimal; If there is a project team burning the midnight oil to meet a deadline which means monthly billing would be more appropriate, then simply switch the affected WorkSpaces over. Once the deadline has been met and the project team have their feet up again, switch back to hourly billing.
End User Devices
The last thing we’ll highlight in this blog for consideration is the end user’s physical device they’ll be using to access the AWS WorkSpace.
Running a full Windows desktop to access the WorkSpace is by far the most expensive option, both licensing and operationally.
With WorkSpace clients available for OSX, iPad, Windows, Android, Chromebook, Fire tablets and simple browser-based access, BYOD is a real option and gives your end users the ultimate flexibility whilst reducing costs. There are obviously a number of security considerations here though, so it’s important the WorkSpace environment is secured appropriately – a topic covered in another blog here: (link to Secure WorkSpaces blog)
So as can be seen, simply deploying and forgetting your AWS WorkSpaces fleet is a sure-fire way to spend more than you need.
AWS have done a great job with the WorkSpaces Cost Optimisation tool, but as you have now seen this is just part of the puzzle; there’s still plenty of work to do in order to get the most benefit.
Contact us if you think we can help.