How to design your VDI infrastructure to avoid boot storms

Solution to VDI BOOT STORMS.jpg
A Solution to VDI Boot Storms by Reduxio

Addressing VDI boot storms has been a gating factor in many virtual desktop implementation projects. Trading off performance and cost seemed to be the only solution: either desktop startup times are severely degraded during boot storm situations, or solid state disk (SSD) solutions are implemented at significant additional cost. The price of SSD-based systems is dropping, but there are other design decisions which can help in implementing a virtual desktop environment that provides cost savings and performance together, limiting the need to provide bad news to stakeholders.

Maybe you've seen some of the resource-constrained performance problems which virtualized environments can experience, and learned your lesson. Every new environment has a tendency to prove its designers wrong, once real users sit down and get to work. Maybe your time is yet to come. Like shoppers rushing the opening of a bargain-laden store, virtual desktops in a VDI environment sometimes rush resources due to demand timing, reflecting user work hours or utility operations such as antivirus scans.

A VDI boot storm is a classic case of a resource rush. Individual instances of a desktop environment join together to consume inordinate amounts of disk reads and CPU cycles while booting their OSes, all on a Monday morning at 8AM. Besides the bad timing, and the required rush from IT staff to address the situation, the storage layer has been traditionally linked to the majority of failed VDI rollouts, due to poor performance or cost tradeoffs.

We’ve seen for a while that having a full, well-performing VDI deployment is achievable. One where you can seamlessly transition between desktops, thin clients and mobile devices; your users will be free to work anytime and anywhere, but we aren’t there yet. Here’s an image from Wikibon on the VDI journey, and as you can see, we all know where we want to go. 

VDI Journey


But the reality is that we aren’t there yet. One common misconception is that if you ‘solve’ the Storage Bottleneck and improve I/O throughput and reduce latency, you’ll be perfectly fine. And this is true to a degree. As newer technologies have been made available and storage performance has greatly improved with the advent of Flash, the reality is that getting a faster storage layer is not the only thing you can do to solve a VDI boot storm challenge.

Modern Architectures = Simple Architectures

Without diving too deeply on the VDI configurations that you can deploy today, it’s enough to say that simplification of your entire IT Stack is much more important that cost.

The more you focus on compressing you stack, reducing the amount of components that you have to deal with the better. The fact is that the fewer number of variables IT has to deal with, the easier and more effective their job becomes.

Storage Optimization via Tiering

While Flash storage is becoming a more viable way for budget-constrained organizations to meet the high IOPS demands seen in VDI environments, All-Flash Arrays are still not within reach of every single organization.

Storage optimization can happen by storing only hot data on flash and the remaining data on high-capacity disk media (HDD). If the active parts of data are stored on flash and the rest on more cost-effective media, you can address the typical compromise that happens between capacity and performance.

However, many systems in the market provide tiering engines that require configuration, policies and/or scheduling. This will only create a new problem to IT staff because you’ll reduce the amount of boot stoom related issues but you’ll increase the workload of admin tasks. So ideally, tiering should be fully-automated and should adapt to changing workloads in real-time to address peak demands during a boot storm.

Until now, vendors either offered systems solely built from a single-flash device type, focusing on high-end, low-capacity use cases, or systems with a limited flash-based cache that only offers marginal gain in system performance and no gain in capacity at all since cached data means duplicated data - one copy in flash and one on disk. Reduxio’s Tier-X™ is an active primary storage tiering technology integrating different types of storage media to create high-performance and cost-effective single, multi-tier storage pool. Data is deduplicated and compressed in-memory on ingest to store only unique blocks to memory and different media types used. Tier-X provides consistent low latency for both reads and writes from or to any storage tier.

Summary: A Solution For VDI Boot Storms

VDI boot storms are a wake-up call for good VDI environment design. Resource performance increases help, but at increased cost, while distributing resource access to avoid competition and demand conflicts at peak utilization times helps make better use of existing disk, network, and CPU investments.

Thanks to Reduxio, you can get all your applications on flash by storing (not caching) hot data on flash.


Watch On Demand Webinar "How to Handle VDI Boot Storms"


Reduxio Systems

Written by Reduxio Systems

There has been no fundamental innovation in data management and protection for primary storage for the last two decades. In 2012, a group of storage industry veterans founded Reduxio with the vision to redefine data management and protection by taking advantage of new processing, networking and media technologies. Reduxio is redefining data management and protection with the world’s first unified primary and secondary storage platform. Based on the patented TimeOS storage operating system, Reduxio provides breakthrough storage efficiency and performance, and the unique ability to recover data to any second, far exceeding anything available on the market today.

Want to comment on this blog post?