Network Computing is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Solving VDI Problems with SSDs and Data Deduplication

It's been amusing to watch numerous storage vendors announce that they, and they alone, have the magic feature (derived from unicorn tears) that will allow you to build a VDI environment to satisfy two cranky constituencies: the users, who will break out torches and pitchforks if their thin-client experience is inferior to their old PCs, and the bean counters, who've drunk the Kool-Aid® and think VDI will let them save capex and opex.

In fact, you don't need unicorn tears to make this work (you can save them for SDN). Data deduplication and solid state drives may be all the magic required.

Storage makes up a significant fraction of the cost of most VDI projects. To help contain storage costs, most VDI implementations use linked clones, which reduce the storage capacity that any given set of virtual desktops will consume.

Linked clones are a great solution for non-persistent desktops, which are the low-hanging fruit of VDI workloads. But once you move past the call centers, school computer labs, hospitals and other environments where non-persistent desktops make sense, linked clones become less and less attractive.

Knowledge workers have a more intimate relationship with their personal computers than people who run one or two applications on a shared PC. They want their Hello Kitty wallpaper and the 300 icons on their desktops, which frankly can be managed as part of their roaming profile or persona. They also want to be able to install new applications once, like the WebEx plugin for their favorite browser, so they don't have to do it every time they join a conference.

You can give users a persistent desktop image with linked clones, but you have to give up most of the advantages of linked clones to do it. The disk space savings fade away as each user's linked clone grows with all the changes from the "golden master" over time.

Consider that in VMware's internal VDI implementation, the average user's clone grows by 1 Gbyte a week. Linked clones typically save 30 to 60 Gbytes per user, but if a clone grows by 1 Gbyte a week, then in about a year your linked clone implementation will actually take more space than if you just created full clones for all the users.

The other big advantage of linked clones is that updates can be posted once to the "golden master," rather than having to be installed on each desktop image. To apply Tuesday's patches, just update the master and recompose the linked clones to include the changes.

The problem is that recomposing the desktops discards the delta file that made each user's linked clone different from the master, so all the users' clones revert to the original state. The users will then have to reinstall the applications, and browser plug-ins, that made their desktop comfortable.

I know of one university where the faculty and staff revolted after the IT group recomposed their desktops and forced the recomposition process to happen only once a semester. Because those systems couldn't go a whole semester without security patches, IT had to go back to installing patches on individual desktops.

Data Dedupe: A Better Way

A better solution is to recognize that linked clones are a primitive mechanism for doing periodic data deduplication, and to replace that mechanism with a more sophisticated deduplication technology in the storage system.

As I discussed in a now-classic post Data Deduplication And SSDs: Two Great Tastes That Taste Great Together, the performance problems deduplication can create on disk storage don't apply to solid state storage.

That means we can use data deduplication to store the common data, like WINSOCK.DLL, across all our desktops just once. All solid-state storage systems with deduplication, like those from Pure Storage and GreenBytes or hybrids from TinTri or Tegile, can dedupe data and still deliver sufficient performance to support thousands of VDI users.

Full clones on deduplicated storage can be managed with the same tools you use to manage your physical desktops and give the users the same rich experience they're used to.

Disclamer: Data Domain (the originator of data deduplication), Greenbytes and Tegile are or have been clients of DeepStorage, LLC.