From the days of the first large-scale virtual desktop infrastructure (VDI) implementations, which began in the 2007/2008 timeframe, VDI experienced a rough start – so rough in fact that VDI almost never happened at all. Early implementations couldn’t meet user demands due to network latency, poor disk performance, lagging graphics technology, and VDI boot storms. Boot storms happened and CIOs around the globe scrapped VDI projects at all stages from planning to production.
A VDI boot storm is the consumption of compute and disk I/O resources during the initial startup of end-user desktop virtual images that results in poor performance for all users. Historically, boot storms and the expense of their recommended remedies are the two main reasons for the failure of VDI projects. (For more on virtualization challenges, see 3 Big Headaches for Virtualization Software Users.)
Patent-pending network and storage technology with compute, virtualization, and SaaS management in ONE enterprise cloud in a box.
Witness the power of Ignite today.
Why Boot Storms Happen
Boot storms occur primarily because storage engineers build storage systems for scalability, redundancy and speed, but no one could foresee VDI’s appetite for disk I/O. VDI implementations aren’t dependent on capacity, scalability and redundancy, or standard speed enhancements or tweaks. VDI environments need read I/O at boot and write I/O during use. First, there’s the read (boot) storm, then the write (login/logon) storm, and finally the steady state write-intensive operations associated with end-user computing. SAN administrators didn’t know what to do. They can adjust storage for read-intensive workloads or write-intensive workloads, but VDI needs both in abundance.
Part of the problem with SAN technology that originally hosted VDI images is that spinning disk technology can’t keep up with the VDI demand for I/O. Remember that SAN engineers can optimize storage for read or for write, but not for both. Only those companies with an extreme need or full coffers could afford SSD technology to alleviate the I/O bottleneck. The rest of the world had to suffer with latency or abandon VDI for traditional desktop computing. (For more on virtualization options, see Faceoff: Virtual Desktop Infrastructures Vs. Virtual Private Networks.)
How to Prevent Boot Storms
It’s easy to prevent boot storms now that technology has finally caught up with end-user expectations of near-instantaneous response times for applications and for desktop functionality. The solutions still aren’t cheap, but they’re more affordable than ever before, plus multiple choices exist to alleviate the performance pains that once plagued every VDI implementation. You can also stack solutions together in order to deliver the best experience to users regardless of device, location, or available network bandwidth.
One of the solutions to the boot storm problem is to issue persistent desktop images, otherwise known as one-to-one desktop deployment. A persistent desktop image means that each user has a single desktop image assigned along with the company’s common applications, any user-specific applications and a desktop profile. Persistence was once the only way to ensure that users had a consistent desktop that remained customized after shutdown and restart. Persistent desktops, when set up for timed “wave” boot up, avoid the boot storms that happen when the majority of users show up for work and initiate their VMs. But as one can imagine, the administrative overhead of persistent desktops is only a little less painful than managing the same number of laptops and desktops.
Technology has evolved to the point where non-persistent desktop images are not only possible, but also preferable. A non-persistent desktop image is one that is a clone of a master image. Each user, upon desktop initiation, receives a copy of the master image that’s coupled with the user’s profile to present the user with the illusion of a custom desktop.
Third-party software companies have come to the rescue in providing solutions to non-persistent desktop boot storms with caching technologies such as image “warm-up” and boot waves.
Caching stores shared image bits in memory for quick access so that only the user’s profile is unique and requires any movement from disk. Common applications also reside in memory with the standard desktop image to further boost initial startup performance. Some companies use an image “warm up,” which is a type of timed caching of boot images. This technology works well in both persistent and in non-persistent VDI implementations. Boot waves are scheduled desktop image startup sequences that prepare desktops for users in waves to prevent boot storms.
The widely acclaimed, but equally expensive, solution is to use flash (SSD) arrays for image booting. SSDs have received considerable attention due to their high IOPS numbers, their durability, their throughput, and their continually falling prices. Enterprise-level flash is now at a price point where it makes sense to consider it as a feasible VDI boot storm solution. SSDs, being 100 percent electronic, provide sustained read and write rates that make the fastest spinning disk drives look like yesterday’s technology by comparison. SSDs are an excellent solution to the VDI boot storm problem with no drawbacks.
VDI implementations are now an accepted and an acceptable form of end-user computing, because the boot storms of yesterday’s VDI have vanished – vanished completely if a company desires to remove them by using technology dollars to do so. The solutions aren’t cheap, but there are many from which to choose. Virtualization vendors and third-party vendors have realized that VDI is a desirable technology and have stepped up to the task of making it work.