How DaaS platforms handle cost optimization in the public cloud
When virtual desktops are running in the cloud, we need to keep a close eye on resource consumption to keep costs down. Here’s how DaaS platforms do it.
With desktop virtualization on the public cloud, there are many ways to be flexible—and in particular, save on resource consumption.
For example, think about being able to automatically scale up instances based on the number of active end users, and then scale down when the VMs are no longer needed. We can also provide VDI sessions with specialized hardware such as a dedicated GPU rented on a per-second basis. Or think about how we can scale from one instance type up to another to get more powerful CPU and more memory and disk performance.
This is very important, because with the public cloud billing model, you pay for exactly the resources you consume, as you go. There are some serious opportunities for cost savings, but you need the integrations in place to make it happen.
Power management integrations
How do we do this? VDI vendors need to have a management service that integrates with APIs provided by public cloud platforms, so that they can power virtual machines on and off as needed.
The VDI platform needs to ensure that it only scales down and powers off machines that don’t have any active user sessions, based on information from the connection broker. Or, when the environment is becoming overutilized, the service needs to automatically power on more VMs and notify the broker that more desktops are available.
The service can also be automated, so instead of being powered on using a schedule, a virtual machine can be started automatically when a user logs on to their client or application portal. This type of automation management also saves more money, since the VDI instances are only running when end users need access to them, but will generate a longer waiting time for the instance to become available to the end users since it needs to be powered on.
Power management in the market
Now let’s look at how VDI vendors are handling this.
First, we have the newcomers in the market, such as Frame and Workspot, which provide fixed T-shirt size virtual machines—i.e., they offer VMs that they consider light, medium, and high-end (with GPU) to end users at a fixed price per month. They tend to have virtual instances with automatic power management, meaning that the instances are powered on when a user clicks on an application. They also have the ability to switch between some instance types (however, you may be limited to a few particular sizes, and not able to pick between every single instance that the cloud provider offers).
Next, we have the larger vendors in the market, Citrix and VMware, which have built tight integrations with the public cloud. Building a VDI environment in the cloud (such as on Microsoft Azure) with Citrix or VMware would means that we need to pay for the cloud compute capacity, in addition to the user licenses for the VDI platform itself.
In terms of provisioning and power management, Citrix has currently gotten the furthest—they now support Amazon, Microsoft Azure, and are working on supporting Google Cloud and Oracle Cloud.
VMware also built their Horizon Cloud offering for Microsoft Azure; for Amazon Web Services, they’ve built their VMware on AWS (VMC) service where we are also able to host Horizon.
Current, both these vendors provide scheduled and automatic power management. This means that they can scale up and down instances based upon load, on a schedule, or a combination, together with user-initiated power on.
Citrix has a service called Citrix Smart Scale, which is included as part of their cloud-based management platform Citrix Cloud, or it can be used separately with a traditional Citrix deployment, which we can then integrate with a public cloud platform for power management.
VMware also has a native service as part of their Horizon Cloud offering, which allows for a similar power management control.
Last, I’ll mention that both vendors support the use of their power management mechanisms for different delivery models, such as persistent and non-persistent VDI
New billing options for management platforms?
Looking at the public clouds where we have pay-as-you-go billing down to the per-second level, I believe it is time for the big VDI vendors to do the same for the licensing models. Most VDI vendors have a licensing model based upon either named or concurrent users, on a monthly basis. Perhaps they should look at their licensing model and adapt for a more flexible way of consuming licenses moving forward?
The smaller vendors provide a fixed cost per user either on a per-hour or per-month basis, which then is combined with the underlying compute capacity and the license, which I believe is the business model that most companies want to move toward.
Moving towards DaaS?
I believe that we will start seeing a big shift toward DaaS moving forward because of this business model.
During Microsoft Ignite, Microsoft announced a new service offering called Windows Virtual Desktop (WVD), which will also provide a mechanism to scale up and down based upon resource usage. This is going to be a free service if you have other Microsoft licenses, but we are still paying for the compute capacity.
Citrix also announced that they are building a DaaS service on top of WVD, and will therefore provide a DaaS offering to customers on a fixed priced like the other smaller vendors in the market.
Most vendors in the VDI market are building tight integrations with the public cloud vendors to be able utilize the scalability and pay-as-you-go billing model. Now, with businesses moving towards more “as-a-service” business models, I reckon that we will see more and more vendors building their own DaaS offering moving forward.
So maybe this isn’t the year of VDI, but the year of DaaS?