ISWest, a regional ISP based 30 miles north of Los Angeles in Agoura Hills, wanted to roll out an IaaS offering to stay competitive in the cut-throat ISP market.
ISWest's business goal was to target existing business-class clients.
Many businesses still shy away from public clouds, and many of ISWest's clients were looking for a way to re-create their existing network environments in the cloud. Typically, these would rely on NAT and use few public IP addresses.
The trouble is that this type of configuration was either not possible with the big cloud providers (AWS, Rackspace, GoGrid), or was complex to configure. ISWest wanted to offer each client their own private network (or multiple private networks) with VPN access to them, and then VLANS would segregate various clients.
After evaluating a range of IaaS providers, including Eucalyptus, Enomaly, OpenNebula and OpenStack, ISWest selected CloudStack because it offered the most turnkey solution. In a bit more than a month, ISWest built an IaaS cloud on CloudStack, and now they have a new service offering to help protect their margins.
Another issue that Swenson of WebMD wanted to address was VM sprawl. "We've calculated that, at minimum, we have 15 percent VM sprawl," he said.
It's easy to understand why. Developers and QA engineers throw something into the cloud, forget about, and then those tests continue to consume resources.
Since CloudStack has an enterprise-ready management interface, it's easy for Swenson to enforce policies. "We encourage experimentation, but now we can set up rules." WebMD gives its developers a thirty-day window to test things out, and during that time, engineers don't need to allocate budget or even prove the project has merit.
However, when these projects are initially set up, they have expiration dates. If the project is a success after thirty days, it's moved out of the test cloud and given its own resources. If the project doesn't work out, it is retired automatically.
This capability also helps keep experiments in house, so WebMD developers use what WebMD has already paid for. "We let people experiment with AWS, but most successful projects come back. We have so much internal capacity that it's just cheaper, and you'll get higher performance, if you keep it in-house," Swenson said.
With VMs reigned in, many organizations want to move next to gaining visibility into infrastructure costs.
"If you jump right to chargebacks, I guarantee you'll experience resistance," Swenson said. "But showbacks are a different story." Showing various business units the VMs, storage, networking and other infrastructure resources attached to their projects changes the economics of them.
Well, the economics don't necessarily change, but the perception of them does. Few will want to tie up resources for a hair-brained scheme that should have been killed off before it ever tied up resources.
"You want it to be a collaborative process," Swenson said. "If you take the time to pull in the VM experts, the network experts, the st
orage ones, etc. and explain how this benefits them, it's not hard to get them to buy into it."
If you manage IT relationships deftly, you can then rein in costs without discouraging innovation.
Both Swenson and Schlesinger emphasized that having an in-house private cloud makes experimentation easier, more cost effective and much easier to manage. In other words, you don't want to discourage people from failing with their tests. You just want failing projects to fail quickly and stop tying up resources.
Jeff Vance is a Santa Monica-based writer. He's the founder of Startup50, a site devoted to emerging tech startups, and he also runs the content marketing firm, Sandstorm Media. Connect with him on Twitter @JWVance.