As enterprises explore private cloud, many look to open source options, such as OpenStack. But they soon discover...
building an OpenStack cloud is no easy task.
Using a managed OpenStack service from companies, such as IBM, Platform 9 Systems, Inc., Mirantis, Red Hat Inc., and Cisco, to name just a few involved in the OpenStack Foundation, removes much of the heavy lifting involved in building, configuring and managing the open source private cloud environment.
SearchCloudComputing.com spoke with Clayton Weise, director of cloud services at Key Information Systems, Inc. (KeyInfo), a colocation provider located in Agoura Hills, Calif., about the company's move into OpenStack and its decision to implement a multivendor private cloud.
SearchCloudComputing: Why did KeyInfo choose OpenStack?
Clayton Weise: I got involved in cloud-based infrastructure and orchestration tools in the early days of OpenStack. [OpenStack version] Dallas had just been released.
[KeyInfo] originally looked at cloud -- prior to our acquisition of the Internet service provider, ISWest -- as a virtualized version of colocation. We needed to differentiate ourselves from large cloud providers. We decided to take this focus on private first/public second approach -- looking at people's concerns with the cloud environment of which performance and security were most prominent.
Our clients have the types of needs where there's a little bit more of a consultative process. We ask them what they are trying to accomplish and we figure out how to leverage some automation and tools like OpenStack to provide that service. For a lot of those organizations, deploying an OpenStack project -- especially in the early days of Grizzly and Havana [versions] -- was a daunting task. It has a lot of moving parts.
One of the things I liked about OpenStack or Cloud.com was the idea of having multiple hypervisors supported within a single framework. There may be a need for VMware, there may be a need for [kernel-based virtual machine] KVM, and we wanted that flexibility.
In terms of something off the shelf and multi-tenant, there really weren't a lot of options. OnApp didn't exist and it's still kind of a niche product. OpenNebula was a little bit further ahead from OpenStack, but you could just tell by looking at the momentum behind OpenStack that it was just going to run right by them.
Was your team skilled in in open source development?
CW: That was actually one of our challenges. We had strong operations skills when we were an [Internet service provider] ISP, but we would use open source -- and we might leverage contractors to do some development work. That's part of the reason why open source is great; you can use it, it's modular and you can mess with it a bit. If you need to, you can bring in a contractor and rewrite some portion of it to make it fit. And, there's no license. That made it pretty affordable.
Were corporate executives or vendors pushing you to adopt OpenStack?
CW: The biggest pressure we got from vendors was using their product as the version of OpenStack versus another vendor. Do you do Mirantis? Do you do Red Hat? Do you do IBM Cloud Manager? Do you do HP Helion? At this point, for a lot of those vendors, this is a bit of a land grab. Each of them is trying to … get as many people into it as possible.
How did you compare vendors?
CW: We didn't pick one; we picked multiple. We have a long background with IBM AS/400 systems, now IBM i, and AIX. We have those in our cloud environment. And we have a large client base that has those deployed to run their databases -- mission-critical ERP systems, logistics or whatever is controlled outside of that. Take a guess which vendor has an OpenStack deployment that supports those proprietary IBM systems? So that was our first direction.
But we knew we weren't going to be limited to just one [OpenStack vendor]. Now we've engaged HP for its system; I've had conversations with pretty much all the major vendors out there -- Platform9, Mirantis, Piston [Cloud Computing, Inc.], Red Hat.
Did you experience any issues getting the OpenStack project up and running?
CW: From concept to getting it up and running in production, it was six months. It's not a quick process in terms of getting the equipment, testing and vetting things out. Part of it is an education process and a switch in mentality. The other big challenge was the networking. The networking model that all this stuff was based on, especially if you're using the predecessor to Neutron [nova-network], was tough to set up. You want to move beyond segregation and have separate services. Both of those took the most work.
What major capabilities does OpenStack still lack?
CW: Until recently, the concept of doing vMotion -- moving a virtual server from one machine to another -- didn't exist in OpenStack. Modularizing the way the network looks [in OpenStack], including integration with other vendors, will really help. Because if an enterprise looks at OpenStack and asks "What do you mean I can't use my F5?" They don't want to have to sacrifice anything.
As a user of both CloudStack and OpenStack, I think CloudStack is a little ahead of OpenStack with regard to being more enterprise savvy. But that gap will close rapidly because there's so much momentum behind OpenStack. Whatever differences may exist between the two projects, I see OpenStack blowing past whatever is out there. Adoption is increasing and vendors are trying to stay ahead of that curve.
This article was edited for clarity.
CloudStack falling behind OpenStack in cloud race
Balancing AWS and OpenStack hybrid cloud