One key advantage is virtualization’s ability to set apart an unstable environment, which is something any developer expects in early phases of application design. As the Microsoft tools are developed, says Friedman, testing early versions may destabilize a developer’s entire computing environment.
“That’s the nature of the beast,” he says. “Almost anything except the simplest desktop application can crash the system. I often tell my developers that if they aren’t crashing the system regularly, they are not trying hard enough. We appreciate virtualization technologies, because they save time, and let our developers spend more of their time on the challenging stuff, not the mundane and extremely time-consuming aspects of prepping test environments.”
Like other virtualization tools, Microsoft Hyper-V lets users ‘snapshot’ the system at a ‘last known good version.’ “We create a rollback that allows us to restore the system to that previous good state within minutes,” Friedman says. “The alternative is having to re-image the computer or re-build the environment, which can take hours,” says Friedman. “This is a tremendous timesaver.”
Friedman’s group also makes extensive use of virtualization and virtual machine images is in quality assurance (QA) testing. “Once you get past unit testing by the developers, we like to talk about the test matrix—what versions of the OS does it run on, for example,” he says. “Then we need to test against both the 32-bit or 64-bit versions of each. Using Virtual Machine (VM) images is the fastest, cheapest, most flexible way to maintain a variety of testing environments. In our QA labs, we are doing this all the time; we do a lot of automated testing, so we stockpile these images as VMs, and point our test suite at them. It’s invaluable. We save so much time you wouldn’t believe it.”
The ability to snapshot, and restore a working image within minutes, rather than wait hours for a rebuilt system, is particularly important. “You’re often looking at bugs where you have to spend a day or two to set up the environment just to reproduce a bug, so the time you spend having to rebuild again and again—that’s a time waster,” says Friedman. “That is loss of productivity.”
For example, when a bug causes a system crash, the developer has to work backward to identify the point where the system crashed. “People often initially set breakpoints too far into the process and the system crashes again. Since a developer can easily be crashing the system several times a day, being able to restore quickly and then re-run the debugging session while the problem is still fresh in their minds is a great benefit,” Friedman says. “When you’re troubleshooting a complicated bug, it’s good to stay in that groove.”
Virtualization is a great boon to iterative testing, developers say. “We have lots of test systems for use by our developers—each of which needs its own x86 control system. Rather than get a lot of PCs, even whatever’s cheapest, we’re using virtualization to create VMs which then ‘attach’ to the external cluster boards,” says Adam Moskowitz, senior software engineer at SiCortex, which builds and sells a range of energy-efficient computers.
SiCortex’s development department uses the free version of VMware Server on a somewhat-beefy rack-mounted server (equipped with 2- or 4 dual-core x86, 4 GB RAM, 300GB hard drive) which can run from four to eight control system VMs, says Moskowitz. “Because there’s very little load on the control system, this works just fine, at lower capital expenditure. Over half of our roughly 25 engineers use these virtualized systems.”
Stands back from the kyebaord in amazement! Thanks!
This is ecxatly what I was looking for. Thanks for writing!