Working on the Client Computing Team at USAA, I got to work on our VDI provisioning automation and farm optimization for capacity and performance. This gave me great exposure to many tools, techniques, and insights both in technology and business.
Before I started, VDI provisioning was a team of 10 doing the work manually with a typical delivery time of 2-4 weeks and with high rates of inaccuracy; when I finished one person part time could deliver more than 100x the machines with high accuracy in just 2 days. This was crucial for projects such as our 32 to 64 bit migration, our migration to a flexible workforce, reducing the onboarding delays, and achieving business satisfaction with the VDI product.
- PowerShell, using the VMware and Citrix snap-ins (before the C#/API clients were available)
- VMware vCenter/ESXi/NSX
- Citrix XenDesktop & XenApp
- Microsoft SCCM + custom internal tooling
- IBM Rational Team Concert (RTC)
Several times we needed to get more capacity quickly for business enablement, and efficiency/performance is of course crucial to user experience and a good ROI. As these came up and in between provisioning projects, I worked on squeezing more performance and capacity out of our systems – both in a hyperconverged Software-Defined Datacenter (SDDC) and in a traditional pod-based architecture datacenter. On just one of the occasions, I found that we could change a storage driver and upgrade the machine type to allow a density increase of 12% while at the same time returning better performance to users. That represented an immediate capacity increase of approximately 5,000 VMs, and the continuous savings is applied as the organization continues to scale up its VDI presence after I left. For some financial context, the hosting for that savings represents approximately $1.2M in hardware/licensing outlays, plus the ongoing support/maintenance/power/cooling.
- Hypervisor/Orchestrator: VMware vCenter/ESXi
- Storage: ScaleIO, XtremIO
- Performance Analysis: SysTrack
- Automation: PowerShell + VMware SnapIns