I got 99 problems, but VDI ain’t one!
We’d be lying if we told you that VDI deployments are easy. Delivering reliable, predictable performance to users tends to be the biggest pain point; so we thought we’d offer some tips we use with our customers that make VDI projects a lot less of a problem. We call these tips the “The CYA Approach to VDI.”
CYA #1: Plan it
- What is the desktop OS you want to use?
- How many virtual desktops do you need to roll out? What’s the right ratio of VM to physical host?
- Can you use the existing ESX environment; or are you building new?
- What are the workloads or applications you want to deploy? How much memory per desktop will that take?
- What’s the average user load and concurrent user load at peak times of the day (the times of day most users will be logging in or out simultaneously, AKA boot storms)?
- What are the CPU, memory, and disk usage stats of the existing physical machines?
CYA #2: Network it
Complaints from users about performance are strongly based on network quality. Latency is usually the biggest reason, as opposed to bandwidth. VDI installations need robust and redundant network topologies; but be warned that the bigger problem is WAN connections to remote clients. WAN traffic really needs to be reduced and optimized, segmented and prioritized, to “enhance the user experience” (stop the whining!).
CYA #3: Secure it
In some regards, VDI enhances security because less critical data is stored on endpoints. On the other hand, whenever resources are shared, security risks are magnified. Runaway processes can affect all users; so enterprise-wide A/V is critical. Virtual desktops are still vulnerable and prone to “ID10T” errors; so they still need to be protected at appropriate levels for the network and the OS.
CYA #4: Work it
- For non-persistent desktops running highly automated apps, such as CRM systems in call centers or retail POS terminals, VDI does well.
- For persistent desktops running basic office apps for documents and spreadsheets, VDI is also good, up to a point.
- For specialized, high-I/O workloads, such as video editing or engineering design: VDI is not ideal (unless of course you can shell out for high-end GPU tech.)
Don’t doom yourself to a painful and expensive future of providing more support and resources over time. To identify if VDI is appropriate, do the following: study the workload, and identify how data associated with the applications are accessed and used over a period of a few typical workdays. If the patterns for I/O and resource demands are low/medium and bursty, you’re good to go with VDI.
CYA #5: Store it
Storage requirements for VDI are markedly different from storage requirements for virtualized servers. While the concept of centralization is the same for storage as it is for infrastructure, I/O and capacity requirements for VDI are different, and probably less straightforward.
Using linked clones and non-persistent desktops based on a single master (or gold) image can dramatically reduce capacity requirements. Likewise we see that inline compression, inline deduplication, thin provisioning and even real-time caching can greatly reduce capacity requirements. However, if these features are not implemented correctly, they can also negatively affect performance.
Minimally, you need performance that can manage the traffic load (including bursts/storms) and concurrent users, the density and the scale. When you understand those user I/O patterns that we mentioned above, you can group similar VDI workloads together in virtual storage pools to segregate storage for different users. The grouping of similar VDI workloads guarantees consistent performance by isolating critical workloads.
CYA #6: Schedule it
Running administrative tasks or “events” in VDI infrastructures is an excellent way to make enemies in a company. VDI management processes, such as creating, allocating, cloning, and refreshing systems and resources all affect performance. Check the settings of virus scans and virus definition updates; and limit full disk scans as much as possible.
Patches, updates, new software rollouts, and backups of storage pools all affect the system differently. These tasks should be performed during off hours or weekends. If that is not feasible, below is a handy chart that lists the I/O impact and frequency of various common VDI events.
Next time, we’ll talk about the ways that flash-based storage from Tegile can help to bulletproof your VDI deployment. In the meantime, if you need to talk about your VDI problems, we’re here to help. Send a message to us at firstname.lastname@example.org.