All-flash arrays might not be the best choice for every VDI deployment, but they're gaining popularity because...
prices are dropping, performance is improving and solid-state drives are growing more sophisticated every day.
Storage often represents the weakest link in the VDI chain, especially when organizations rely solely on hard-disk drive (HDD) systems. Simply put, spinning disks cannot keep up with the demands of hundreds, or even thousands, of virtual desktops running simultaneously.
VDI storage challenges
The highly random IOPS, fluctuating between the periods of read-heavy and write-heavy high-density loads that come with VDI, strain storage capabilities beyond what many HDD systems can handle.
Boot storms, for example, occur when thousands of users log in to their systems each morning at the same time. They force storage disks to contend with an influx of simultaneous read operations that can stretch the limits of even the staunchest HDDs.
VDI storage must also contend with read-intensive periods during antivirus scans, as well as the heavy writes from software updates and patching. Then there are the unpredictable I/O patterns that occur throughout the day just from users all working at the same time. Even simultaneous logoffs at the end of the day can overwhelm the storage system.
VDI storage options
One alternative to HDDs is server-side caching with flash. In this approach, a flash device connects to the server's Peripheral Component Interconnect bus, where it caches the data users access the most to reduce network loads and I/O processing. For server-side caching to be effective, however, the caching algorithms must accurately determine which data to cache.
Admins can also implement storage-side caching on the front end of an existing HDD system. This method uses flash to maintain a nonpersistent copy of the most active data, participating in both read and write operations.
Another option is a hybrid array, which mixes solid-state drives (SSDs) with HDDs to address the challenges of VDI workloads without fully investing in all-flash. Admins can use their existing hardware and still get some of the benefits of a flash approach. With hybrid arrays, admins have to make two very different systems work together, and they still face the limitations HDDs.
Not surprisingly, the gold standard of flash-based storage is the all-flash array, which eliminates the HDDs altogether. An all-flash array can handle the fluctuating I/O streams and heavy read and write operations typical of VDI workloads.
Moving to an all-flash array
For larger VDI implementations, all-flash can represent a significant boost in performance and a worthy investment. The SSD array can deliver the IOPS and capacity and handle the latency-sensitive workloads common to VDI deployments with speeds other options can't match.
An all-flash array can manage boot storms, virus scans and software updates without a problem. It can also support more desktops per disk than other approaches, alleviate I/O bottlenecks and deliver performance users can appreciate. Plus, all-flash arrays require less space, power and cooling resources to maintain.
The latest generation of all-flash arrays delivers greater capacities. Storage products that once supported hundreds of gigabytes now handle more than a petabyte of data. The single-level cell array, which stores one bit of data per cell, has given way to the multi-level cell and triple-level cell (TLC) arrays, which store two or more bits per cell.
Vendors also deliver their all-flash arrays with sophisticated compression and inline deduplication capabilities that can increase capacity even more. The compression features remove redundant data within a file, and the deduplication features remove redundant data across the files. In addition, many all-flash arrays now include optimization technologies, such as continuous data protection and guaranteed quality of service, as well as disaster recovery features, such as the ability to generate snapshots.
The other side of the all-flash story
All-flash storage is still expensive, and a good percentage of organizations don't really need to go the all-flash route.
Shops supporting massive VDI implementations, however, cannot ignore the challenges they face with their current storage set up. An all-flash array might be the only approach that can guarantee the performance they need.
HDD arrays still come in at a lower per-GB cost than all-flash arrays, but the differences are shrinking as vendors produce more flash-based products. In addition, flash's smaller footprint can result in savings.
Technologies such as data compression and deduplication can offset the costs even further. To address the nuances of cost comparisons, many all-flash pundits suggest admins take a per-IOPS approach, which more accurately reflects the VDI workload.
Even an IOPS-based approach can be difficult to assess because flash storage becomes less reliable as the chip gets more dense. The more dense the disk -- the TLC array -- the faster the cells wear out, requiring a certain level of redundancy to protect data.
When it comes to pricing VDI storage options, especially when comparing all-flash to HDD, admins must consider all the variables that play a role in determining the true costs -- not only the initial capital outlay, but also the expenses that come with long-term support and management.
Is all-flash the answer?
Before making the decision to go with all-flash, admins must be sure it is indeed what they need. And they must take into account all the variables that go into VDI storage, such as the type of workloads they support, how they piece together their infrastructure and the number of users and their work patterns. Above all, admins must define their performance goals and determine which alternatives meet these goals.
How to approach persistent VDI storage
VDI storage buyer's guide
A closer look at VDI storage requirements