Boot storms are one of the biggest problems VDI administrators must cope with, because they can drag out the virtual...
desktop login process and hinder VDI session usability.
A VDI boot storm refers to a large number of users logging in to virtual desktops at the same time, which causes a decrease in virtual desktop performance.
To put an end to boot storms, IT must understand why they happen in the first place. Insufficient storage I/O makes it hard to deliver data expediently over the company network. That's often the cause of a VDI boot storm, but it isn't the only potential issue. Simply put, boot storms happen because at least one VDI resource -- storage IOPS or network bandwidth, for example -- cannot keep pace with the demand when everyone logs in at once.
The first step in addressing boot storms is to use performance monitoring to establish the cause of the problem. In most cases, the only way to eliminate boot storms is to upgrade the hardware component causing the performance logjam, but a few other temporary fixes can also mitigate the effects of boot storms.
Batch booting before workers arrive
The most effective way to decrease the effect of boot storms without investing in additional hardware is to power up virtual desktops in batches. By having some virtual desktops running and ready to use when employees arrive at work, IT decreases the number of desktops users power up when they arrive, thereby decreasing resource consumption.
Batch booting is not a perfect fix, though. If IT powered on all virtual desktops prior to the first user arriving for work, it could effectively eliminate boot storms, but it would still have to contend with login storms. Those situations occur when a large number of users access the operating system at the same time and, similarly, resource demand spikes from loading up users' applications.
When network bandwidth is to blame
If IT determines that the network causes the performance bottleneck, it should consider using network interface card (NIC) teaming. That process links NICs to a single virtual switch and allows multiple physical network adapters to work as a team. The aggregate bandwidth is greater than what a single NIC provides on its own, offering more juice to deliver users' virtual desktops during busy login periods.
IT should also check if it can offload any traffic from the NIC team. For example, if IT can move management traffic to a different NIC, it can free up bandwidth for virtual desktops and improve their performance. Some NICs support various forms of TCP/IP offloading, which can improve storage performance depending on the product admins use.
If storage IOPS cause the VDI boot storm
If boot storms are related to insufficient storage IOPS, then IT has a few options to fix the problem.
First, it can implement solid-state drive (SSD) caching, which places frequently accessed data on SSD drive to accelerate storage delivery to end users. SSD caching typically offers a noticeable improvement in storage performance. Just keep in mind that VDI storage demands tend to be roughly 30% read operations, which retrieve data, and 70% write operations, which send data to storage, although these numbers can vary widely from company to company. As such, it's a better idea to use a read/write cache instead of just a read cache.
IT can also fine tune its virtual desktops to minimize storage IOPS. For example, many organizations configure their antivirus software to scan write operations but not read operations. Another way to decrease storage IOPS is to disable pagefile usage for virtual machines. A pagefile is an area of a hard disk reserved for RAM data that is not used often. Whether VDI shops should uses pagefiles on virtual desktops has been a hot debate for years, and admins should consider their usage on a case-by-case basis. If IT decides not to use pagefiles, it must allocate plenty of memory to virtual desktops to ensure that there are enough resources to prevent boot storms.
Finally, use a storage configuration conducive to virtual desktop performance. Admins should configure storage arrays in Redundant Array of Independent Disks 10 format, or RAID, so there are copies of data on multiple hard disks. Furthermore, they should not use dynamically expanding virtual hard disks, because the expand-on-write process produces considerable storage overhead.
Storage or network hardware upgrades are the most effective way of mitigating the effects of boot storms. But if new hardware isn't an option, then the best thing IT can do is use these guideline to minimize resource contention.
Prevent performance issues from boot storms
Stop boot storms with a solid-state drive
Using SSD to improve VDI performance