Jaymes Davis, Senior Director of Innovation and Professional Services Organization, Office of the CTO, Sphere 3D
The time for Distributed Desktop Hyperconvergence (DDH) is now. Rather than being based on a data center-only model where server virtualization presents limitations, DDH will address desktop virtualization in a decentralized architecture. DDH uses a different technology so that such server virtualization limitations literally do not exist, while retaining the most useful features of server hardware and virtualization technology. Such as GPU virtualization and Higher Core counts are increasing desktop consolidation. But for DDH to be effective, it will need to combine the flexibility, performance and local compute power of a physical desktop with the advantages of virtualization such as high user to admin density.
DDH when residing in data centers and branch offices will cover 100% of the workforce and will rely on less expensive and more plentiful desktop administrators and help desk support, rather than centralized IT departments. In order for this to be successful, therefore, the workflow will need to be the same as for physical desktops. DDH must be able to scale up within one location and scale out across multiple locations (scale out is technically difficult with traditional data center-only architectures).
DDH must support virtual desktop failover and recovery from one machine to another anywhere in the world, and it must do so automatically and near-instantaneously. This is the opposite of today’s data center-based technology which can only fail over within a site itself, and is often a manual process that takes hours if not days to complete.
How can we build a Distributed Desktop Architecture? Sphere 3D first created Desktop Cloud Orchestrator (DCO) to manage a distributed desktop architecture from a logical and physical perspective). DCO provides admins with the ability to manage down to the user a wide range of tasks ranging from power operations to recomposing an image. This leaves the enterprise architects free to focus only on managing the virtual environment. DCO can fail over and recover from a site to another site far away allowing scale out. It can do this automatically and instantaneously so users do not suffer down time.
After Sphere 3D solved the fundamental technology problems faced by VDI, it then moved on to solving business issues. DDH must be simple enough to plan and deploy faster than physical desktops. What’s the point of a productivity enhancer if it takes 6-12 months to deploy? Businesses often change radically in that length of time.
Sphere 3D introduced is behind the rationalized compute platform. Here the V3 Platform of three different appliance lines is very important. Unlike traditional data center-only VDI, V3 focused on the end user experience vs. the server admin experience. V3 expanded beyond just Knowledge Workers to provide two versions for light and heavy graphics users. Traditional data center-only VDI requires significant analysis and study of existing end user computing usage to determine the specs of the VDI deployment. The reason is that fluctuating usage can significantly change performance from day to day or even hour to hour for all users. As a point of fact, most traditional VDI deployments just add a huge resource cushion. This heavy analysis and extra resource are no longer needed with the V3 RCP appliances. Simply add the users who meet a certain profile and find the V3 appliance or appliances that can handle that demand. A Sphere 3D channel partner can size, purchase and deploy a V3 installation in less time than it takes to order physical desktops from a manufacturer.
Distributed Desktop Architecture – The Vision
A distributed virtual desktop appliance that can be sized to fit the user workload and the number of users per site really changes the power, scope and reach of what can be achieved for a client. It requires no change in the enterprise’s IT operations and significantly improves the economics.
An example is our V3 Flex Nodes that deliver containerized desktops in a 1U appliance to 10-100 users, yet without a change in hardware can scale out across multiple locations to 10,000 users. So now, a business with 1,000 locations with 10 users in each location can cost-effectively deploy and manage VDI. Or, Sphere 3D’s End User Computing Nodes for Knowledge Worker and Business Graphics can be used to provide graphics capabilities to select users on an as-needed basis. The current solution using data center-only VDI does not support high graphics so instead each user must be allocated an expensive physical machine. Since most business graphics users only use graphics for part of the time, that expensive hardware is mostly under-utilized, representing a significant wastage in resources. Most importantly, both Sphere 3D’s Flex and End User Computing (EUC) appliances can be used across the WAN for failover and Recover from site to site. These are use cases that are impossible to replicate using traditional data center-only VDI technology.
As an example of DDH from an architecture perspective, imagine you have 40 users per office and 100 offices across the globe. If we look at a centralized data center architecture, this would require servers representing 4000 users in one location. This means that there is one single point of failure for all users from the WAN perspective. To plan for logical failure of servers, one must plan for over-capacity to support the N+1 design to have service during an outage. The current Recovery Time Objective (RTO) to bring up 4000 desktops from an Input Output Operations Per Second (IOPS) perspective will require time.
Conversely let’s create an architecture using DDH where each office has a 1 knowledge worker appliance (an EUC node that supports 60 users). There is no single point of failure as the appliance can fail over to another site. 20 sites would have to go down before the end-user saw a service interruption. If local Internet went down at a site, local users could use the local appliance to keep working. By distributing workloads to the edge, we can reduce server sprawl within the data center and the resilience of ROBO locations.
Currently, we are at the point where “The year of VDI” has come and past and we are being optimistic about the growth of VDI within our environment. PC sales are on a slow decline of 3-6% per year depending on which analysts you believe, and there is an increase in both mobile devices and thin clients of 30-60% (again analyst dependent). This information presents what I call the “gap of reality.” PC utilization is still prominent in enterprises today. Thin clients and mobile devices are being used to supplement the desktop. The desktop will be here for years.
Sphere 3D’s Distributed Desktop Hyperconvergence vision is to match form and function for EUC placing compute in a shape that matches the user base by size, workload and location. The aim is to match your VDI management with the people you already have managing desktops. It is to match changes in your virtual infrastructure to changes in your business. Above all, it’s to give your business the power to determine your technology infrastructure, not the other way around.