Hybrid Desktop Virtualization: A New Approach for the Cloud - Page 2

Issy Ben-Shaul (Profile)
Wednesday, May 11th 2011

Because hybrid desktop virtualization leverages the compute power of the physical endpoint, it is far more scalable than VDI, supporting thousands of endpoints per server node, compared to dozens of endpoints in a VDI environment, hence requiring a fraction of the cost of VDI in datacenter infrastructure. Because servers mostly manage and store images but do not execute client workloads, coupled with the fact that clients can continue to operate when the server is not available, this model lends itself to stateless design and scalable clustering for both scalability and high availability. 

As noted earlier, the primary copy of each desktop resides and is managed in the data center inside the private or public cloud. Snapshots of the centrally-stored desktop can be used for two purposes. They allow users to restore deleted, corrupted, or overwritten files from the PC client, without a call to the helpdesk. Snapshots also enable administrators to quickly restore the complete desktop to a last-known configuration state to the same or a different endpoint, in the case of any corruption or malfunction of the original endpoint.

A Better Way to Handle Software Updates

While the hybrid model can work well for many smaller and mid-sized organizations, enterprises with many branch offices may find that distributing significant software updates or large images to each client in each branch can require hefty bandwidth and dedicated image distribution hardware at each branch. Some solutions solve this problem by enabling IT to designate one or more branch desktops as a software proxy. This allows IT to do a service pack update, for example, by upgrading the central image then downloading the new image once across the WAN to the proxy desktop. All other endpoints whose desktop is linked to that image receive the update locally from their branch proxy, significantly reducing WAN traffic.

Many user and system files are identical or similar across users, leading to large redundancies within and across multiple desktops at both the file and sub-file levels. Advanced network deduplication technology can eliminate this redundancy and thus further reduce the amount of data that needs to be transferred between the data center and desktops. For example, 3,000 endpoints may have the same copy of a PowerPoint file, but only one copy should be transferred (and stored). Once the first copy resides on the server, there are solutions that will recognize that the file already exists on the server database and will not transfer or store the file again. Further, if changes are made to a file, only the changes are uploaded or downloaded, not the entire file.

Finally, any changes (upgrades, new apps, etc) that are made by the IT administrator to the central base image propagate to the endpoint in pre-installed form and merge with user layers, eliminating the need to install packages at the endpoint.

Lost Laptop? No problem

Because hybrid desktop virtualization systems work within the OS, they have the unique ability to determine which files are absolutely critical for a desktop to boot. Since this is just a small subset of the total desktop size, these few files can be transferred to the endpoint and used for the reboot. The remainder of the user data and files can be streamed in the background later, in order of the frequency they are typically used, once the PC is up and running. If a user requests a file before it has been transferred, the server can send the file with high priority. This allows users to return to productivity in minutes instead of hours or days. These features, as well as snapshots, can also be used for disaster recovery if the endpoint becomes corrupted.