This crossed my path this morning, and I wanted to share. A hot-off-the presses reference architecture doc co-authored by EMC, Mirantis and Canonical (Ubuntu of course) detailing HOWTOs for two of the key OpenStack partners and distributions.
The Mirantis one covers the most common 3 EMC Openstack Cinder targets (VNX, XtremIO, ScaleIO) and Mirantis Openstack (Juno) including Fuel. The Ubuntu one likewise covers VNX and XtremIO and Ubuntu JuJu.
Click below to get the Mirantis and Canonical docs.
There’s a lot going on in the OpenStack space, and a lot of internal dialog I see amongst the EMC SE teams on the topic of Openstack. This intentionally provocative post on El Reg (here) based on a Gartner comments at a recent summit in Sydney was a bit of a hot-spot.
Here are my personal perspectives (through my own eyes/experiences – so take it with a grain of salt, I don’t claim to be a savant):
Q: OpenStack being used in the enterprise? A: Yes.
Q: Does it take skills to deploy and maintain? A: Yes, and I would argue more “ponytails” (aka people with core Linux skills and willing to understand code – but not necessarily developers) than most “Enterprise products”. Companies like Mirantis and Canonical enhance and work to simplify OpenStack (see Mirantis Fuel and Ubuntu JuJu as examples) – but it’s still not easy to maintain (it’s pretty easy to get going).
Q: Where is it working best? A: Where people don’t press it to be something it’s not (a substitute for vSphere and the vRealize stack for workloads that demand rich infrastructure services). Don’t expect it to have the degree of maturity one would expect of a “product” for the newer elements of the full OpenStack framework (higher-order management/orchestration/reporting) – but of course, you can do anything given the skills.
Q: If you don’t have those skills, should you partner with people who do (aka services oriented businesses that also have value-added distros like Mirantis and others)? A: Yes.
Q: Are there alternatives if you don’t want to go the services route? A: Yes. There are numerous parties out there that are pursuing the other route (“industrialize” Openstack – even so far as to package it as a managed service) – examples like BlueBox. It’s the VMware intent to “industrialize” Openstack via VIO. EMC’s intent is to support VIO as well – but also Project Caspian indicates that we think there is also another “industrialized” appliance/managed service route as well.
Q: Do “services” represent “lock-in” (one of the claims made in the article). A: Crikey – what hog wash. Can we start to focus on “lock-in” where it really applies? Lock-in exists anywhere you make a decision. The question of “degree of lock-in” is a function of the opportunity cost/cost of displacement inherent in a decision. The cost of displacement of the whole world of infrastructure (right up to the IaaS layer) is dropping to be negligible relative to the choices of SaaS/PaaS/Data Fabric elements. This bogeyman of “lock-in” is getting laughable. Get used to making decisions :-)