Virtual desktop infrastructure (VDI) is well on its way to becoming the next big trend in virtualization in enterprises but before that can happen, a viable back end storage solution must emerge that can cost-effectively deliver VDI’s conflicting capacity, cost and performance demands. One specific configuration that end-users find particularly appealing are those storage solutions that deliver both deduplication and SSDs.
Ever since end users transitioned from using dumb terminals with command line interfaces to PCs with full-featured GUIs, there has been no looking back. However the transition to PCs also introduced the era of distributed data, difficult to support desktops and insecure environments – all of which corporate IT was stuck supporting. Virtual desktop infrastructure (VDI) promises to restore some semblance of order but before VDI can be implemented a back-end storage solution that accounts for VDI’s sensitivity to storage capacity, storage costs and storage performance demands is needed.
VDI technology now offers a Windows GUI like end-user experience with the security, simplicity of management and centralized data storage once only found in main frame environments with dumb terminals. Couple this technology with the availability of inexpensive hardware components at the desktop, networking and storage layers and VDI has become both affordable and practical for all.
This has led to high levels of interest in VDI. A 2009 survey by Imprivata, Inc, of executives in 132 organizations found that 73% were investigating or had plans to implement VDI in 2009. Of these, 78% cited simplifying desktop management as their primary reason for their interest in VDI.
But as organizations look to adopt VDI they are finding that on the backend they need storage that scales to high levels of capacity and performance at a low cost per GB.
A prime example are those enterprises that are looking to virtualize hundreds or even thousands desktops. Each virtualized desktop needs about 20 GBs of useable storage capacity creating the following storage capacity requirements:
# of Desktops VDI Storage Capacity Needed
10 200 GBs
100 2 TBs
1,000 20 TBs
10,000 200 TBs
100,000 2 PBs
Virtualizing these desktops also creates new performance demands that do not exist in desktop environments. When users come to work boot-storms occur when hundreds or thousands of users log on at the same time that can result in read/write ratios of 20:1 or higher. Then once these boot-storms subside, I/O activity will swing more heavily towards writes with write I/O activity accounting for 50 – 80% of the total I/O activity the rest of the time.
So while high end storage systems may deliver on these capacity and performance demands, they will cost much more on a per GB basis than enterprises can typically afford. Conversely, entry level storage systems may come in at the right price per GB but will fail to meet VDI’s scalability and performance requirements. So what VDI environments ideally need is a storage system design that both scales and performs without keeping the storage costs per virtual desktop on par with that of a desktop hard drive.
In that vein some techniques have already emerged that deal with these competing needs that VDI has. For instance, it is widely recognized that most desktop data is duplicate data so when desktops are virtualized it makes it both practical and possible to deduplicate this data.
Deduplication then has the potential to dramatically reduce VDI’s storage capacity requirements while simultaneously improving its performance since deduplicated VDI data can be put on high performance solid state drives (SSDs.)
Granted, SSDs are normally cost prohibitive. But if data deduplication ratios reach 10:1, 50:1 or even higher which is possible, the cost of using SSDs to store deduplicated data can drive down the cost of using SSDs to a level that is even less than what desktop HDDs cost on a per virtual desktop level.
Making this combination of using deduplication and SSDs even more intriguing is that performance for virtual desktops in these environments could literally skyrocket. Access time for reading data stored on SSDs is measured in microseconds (as opposed to milliseconds when accessing data stored on desktop HDDs.) So users accustomed to waiting a few minutes for their desktops to come online may now only need to wait a few seconds.
It is this combination of faster performance and lower costs that have many excited about what the future holds for VDI and why, on the surface, deploying this type of storage solution as part of a VDI deployment almost sounds like a no-brainer.
Unfortunately it is not quite that simple. Both deduplication and SSDs have their own sets of drawbacks that must be accounted for and managed before they can be safely deployed into a VDI environment. It is these risks and what actions the storage solution should be taking to mitigate them that I will cover in an upcoming blog entry.