Virtual Desktop Infrastructure (VDI) is becoming more and more prevalent in larger IT shops as a way to reduce client-side costs. That is a good thing for most organizations, but what they are finding is that the underlying infrastructure used to support them becomes increasingly important as the number of VDI clients grows.
One result: the availability of I/O bandwidth can become critical at peak times as hundreds of clients all start up at the same time in the morning.
The good news is that the solution to the problem can be fairly simple. Some of the test results I have seen for solid-state disks (SSD) show that they could be used to solve the I/O issue. SSD runs at memory speed, so it has the potential to provide a huge performance advantage over even high-end FC or SAS drives.
In fact, an SSD drive can be used as a type of high-speed disk cache for applications that require a temporary or permanant speed boost.
The problem with SSD is that it is still fairly expensive, so most organizations cannot afford to simply replace existing storage resources with it. SSD is usually added on to existing storage arrays (most new arrays now support SSD drives), provided via an appliance into existing SAN infrastructure, or added internally into the servers themselves via a high speed PCI slot.
Since there is usually a finite amount of expensive SSD available, the problem is how to most effectively allocate and share those costly storage resources.
Standard SAN resource allocation techniques do not
provide an effective means of optimizing the allocation of that pricey storage for applications when the need arises, such as the VDI boot storm in the morning or month-end database batch proccessing. Also, those expensive disk resources need to be shared across all applications for them to provide the most cost benefit to the organization.
With standard SAN LUN allocations methods, if a storage LUN is assigned to application A, it cannot be used by application
B unless those servers are clustered together. Therefore, if a mission-critical database server needs use of the expensive
SSD resources during batch processing, they cannot also be dynamically allocated to the VDI infrastructure server to handle
the morning rush.
The benefits of intelligent storage virtualization -- such as the creation of a physical abstraction layer between servers and storage, host transparent data movement between storage pools, and performance-based I/O policy creation (based on the
performance requirements of the application) -- become apparent in solving this problem for VDI infrastructure.
Storage virtualization creates pools of storage with each pool providing different properties and methods that can be dynamically allocated to different applications.
Storage pool properties can include:
Using an intelligent storage virtualization solution, the SSD resources can be provided as a tier-one resource for the VDI infrastructure in the morning, and then freed up and dynamically reallocated for the database servers in the afternoon.
The ability of the virtualization solution to create policy-based thresholds for throughput and performance enables the creation of watermark policies for transparent movement of data between SAN tiers.
My own internal testing shows that when used in conjunction with storage virtualization with dynamic data movement, the addition of one percent (1 percent) of SSD to an existing SAN can provide a one hundered percent(100 percent) performance boost or applications.
Now THATS a good return on investment!