Also known as boot storms or VDI boot storms, cloud storms occur when a large number of users attempt to boot a cloud service on their computers in the same time frame, degrading the service. The network becomes overloaded with data requests. When companies provide virtual desktops to their employees, it’s common for a large number of virtual machines (VM) to be requested at the beginning of each work day. Cloud storms can also occur during a virus scan or when a system update forces a system-wide reboot.
When virtual desktop infrastructure (VDI) environments first came out, storage companies were concerned about cloud storms and how they would have an effect on system performance. Rightfully so, a cloud storm can be harmful to the performance of your virtual desktop infrastructure and is an extremely stressful situation for cloud hosting providers and your storage. It can also cause a drag on network throughput and storage I/O.
Although a cloud storm can be devastating to an unprepared VDI environment, its resources and productivity, it can be prevented. Choosing the right VDI solutions with the right architecture is a great place to start.
One simple way of preventing a boot storm is staggering the powering-on of your VMs. This is an easy solution when you have a server virtualization environment with VMs that often have 99.99% up-time and remain powered on almost constantly. However, this strategy requires you to control when your end users decide to power on a desktop VM, and that can be a trying task.
Throwing more storage at these storms might sound like the answer to the issue, but it’s often not enough and can be costly. Instead of replacing your existing storage system, you can solve a boot storm with a third-party, cache-based solid-state solution, like AWS ElastiCache and Oracle RAC’s Cache Fusion. These solutions have the advantage of broad use, and they can be used for other things like a busy Oracle application once the boot storm is over. Solid-state drives (SSD) will allow you to take on the high I/O that occurs and it is much more cost effective than outfitting your whole storage array. These third-party solutions will allow highly scalable applications to be built without partitioning.
If you have any questions regarding cloud storms, or anything else related to your network capacity, please contact us.