Kaviza is the third vendor to offer a submission to the Virtual Desktop NAS Vendor Challenge. Today’s guest post comes from Kumar K. Goswami, Kaviza Co-Founder & CEO. Kumar’s guest post begins below.

—–

Your  notion that many customers are just priced out of desktop virtualization and that much of the cost and complexity with VDI (sever-based desktop virtualization) has to do with it being layered on top of server virtualization was precisely the driving factor behind the genesis of Kaviza.  Our mission has been to drive down the cost & complexity of VDI and provide “VDI for the rest of us”.  What is even uncannier is the similarity in names.  You call your notion “virtualization infrastructure in a box”.  We call our product we launched recently, “VDI-in-a-boxTM”.

The philosophy behind our approach is as follows:

1. Make the solution extremely simple.  Package everything as an appliance, load it on a server with a hypervisor and that’s it — you now have a fully functional, self-contained virtual desktop server appliance with everything you need to manage templates, create, provision, load balance desktops and login users.
2. Make the infrastructure cost effective and require nothing other than commodity servers with direct attached storage to manage and provision the desktops. That means no shared storage or high speed interconnects that jack up costs and cause central bottlenecks.
3. Ensure that the system can be grown on-demand easily without requiring lots of manual activity, capacity planning and so on.
4. Provide a higher level of abstraction where the management is about desktops, users, templates (golden images) — stuff the desktop IT staff cares about as opposed to virtual machines, server pools and virtualization details.
5. Don’t re-invent the wheel and leverage best of breed components seamlessly. We are protocol agnostic and hypervisor agnostic.  We leverage Active Directory/LDAP for user management.  We tie in with application streaming solutions like AppV seamlessly (for those who want it) and we can work with Active Directory’s roaming profiles or personalization modules from AppSense or RTO for those who want user personalization.

How did we do it and what’s the architecture?

As shown in the figure below our solution is distributed and consists of one or more servers (you need at least two if you want high-availability otherwise you can use just one) each running a hypervisor and our Kaviza Manager (aka kMGR) virtual appliance.

The kMGR appliances on each server communicate and work together to

• Run the desktops
• Ensure there are redundant copies of key data so there’s no single point of failure
• Dynamically and automatically incorporate new servers
• Detect and dynamically recover from server failures
• Simplify management by allowing the administrator to manage the solution as if it were one logical server

The figure below shows the key modules in the kMGR virtual appliance that provide the above functionality:

They are:

• Grid engine:  This module communicates with all the other kMGR’s grid engine to ensure that there is a cohesive, coherent grid of servers.  It manages the communication and ensures there is ONE global notion of the state of the grid.  The grid engine module creates a hot-pluggable grid and enables servers to be added and subtracted on demand.  Servers are added by simply answering 2 questions and providing authentication to join the grid.  The kMGR then ensures the newly added server is provided with all the needed configuration and template information to participate in the grid.  Similarly, when a server is removed or fails, the grid engine detects the missing server and automatically ensures that other servers take up the slack.

• Logical shared storage: This module ensures that all key information such as user information, desktop configuration information and the templates (the golden images from which desktops are created) are copied to other servers in the grid to ensure that there is no single point of failure.

• Load balancer: This module load balances the desktops across the grid to ensure optimal use of the grid resources.

• Template management: This module provides the tools to manage the lifecycle of the templates which contain golden images of the OS and application, CPU and memory specification of the desktops created from it and policies that dictate when desktops are regenerated from its template.   The template module uses “linked clones” where multiple desktops are generated from a base golden image to save storage.

• User management: This module ties in with Active Directory or an LDAP server and ensures that users have authorization to use a desktop and manages all user sessions.

• Provisioning engine: This module works with the others and does the detailed work of provisioning and generating desktops across the grid based on authorization and policies set by the administrator and directives from the template load balancing modules. Administrators do not have to manually provision, load balance or manage capacity.