VFD4 Prep: Platform9

Platform9 bill themselves as “the simplest way for Enterprises to implement an OpenStack based, agile, self-service Private Cloud.” Ok.

They’re offering a Software-as-a-Service OpenStack Controller. It looks like they have some fancy software that lives somewhere out there on the Internet, and talks to your infrastructure over HTTPS. You sign up for an account, and then manage your infrastructure using their cloud service.

I guess it layers an OpenStack API and web-based service catalogue in front of your infrastructure to make it easy to provision and control, much without all the finicky challenges of setting up your own OpenStack environment.

OpenStack is notoriously complex at the moment, so efforts to make it easy to use will definitely appeal to companies who just want the benefits of private-cloud without all the hassles of setting it up.

Let’s see what’s needed on their How It Works page.

You need an agent, deploying it on your Linux servers. Checking the Supported Platforms page, we have Linux/KVM support only, for CentOS and Ubuntu for both hypervisor and guest OS. This looks very OpenStack focussed, which shouldn’t really be a surprise.

Support for vSphere is apparently coming, as is Docker support. There’s no mention of Hyper-V.

Confusingly, the page about Why Platform9 has copy (at time of writing) that says it “Fully supports KVM, vSphere, and Docker” which it doesn’t (yet). It also says it “Supports most popular Operating Systems” so long as they’re CentOS or Ubuntu? Or do they mean Windows guest provisioning is ok too?

Update: Platform9 updated their website as a result of this article. They’ve clarified when KVM, vSphere, and Docker will be supported. They’ve also updated the supported operating systems list, which now includes RHEL for hypervisor, and guests include RHEL, Windows Server, and Windows 7/8. All 64bit, so I don’t know if no 32bit is just an oversight or what.

Anyway, kudos to Platform9 for updating official web copy so quickly in response to feedback!

Making Private Cloud Easy

Platform9’s target audience, according to the datasheet (you can download it from here), is twofold. Quoting from the document:

• IT and dev/test leaders looking to empower their users with agile, self-service provisioning
• Organizations looking for a private alternative to AWS

I think this is a poorly phrased way of saying what they really want to do: make private cloud easy.

People who want private cloud want it because of two major reasons. Firstly, they want the speed, visibility, and elasticity of AWS. Traditional IT takes a month of Sundays to stand up a server, it costs too much (i.e. people don’t understand the value being provided, which is often less than the cost to be perfectly honest) and it’s hard to turn off when you don’t need it any more. It’s hard to move around, and hard to buy more.

Secondly, they want it within their own data-centres for control reasons, so they want something that works like AWS, but that isn’t AWS. They’re not comfortable with putting all their eggs into a public cloud, or prevented from doing so because of compliance reasons.

But building your own cloud is hard. AWS took years of effort by lots of very smart people, and it only very recently added enterprise features like audit logging. VMware has a rich ecosystem and is the de-facto standard for virtualisation in enterprise, but it’s not exactly cheap. And look at the cachet accorded to those with a VCDX. It’s a tacit admission that doing good virtualisation designs (and implementing them) isn’t easy. OpenStack is even harder to stand up because it lacks many of the robust features of something like VMware, or they’re relatively new. Check out the guide to HA on OpenStack for an idea of the complexity.

The appeal of a “just turn it on” private cloud is pretty clear, but if everyone else has trouble doing it well, how does Platform9 do it, and do they do it well?

Questions for Platform9

Here are some questions I have about the Platform9 solution:

Why SaaS rather than a virtual appliance? Why do I have to expose the control plane for my infrastructure over the Internet when Heartbleed and Poodle are so fresh in people’s minds? Why can’t I have my central control point secured behind firewalls in my own environment?

How do you co-exist with existing infrastructure, tools and processes? You say it all just happens, but how does it do this well?

Much is made of developers, but what about operations? Does this now control my production as well as dev/test? If I don’t move everything to Platform9, what are the implications transitioning an application from dev/test to prod?

How does the service design part work? Do you have a set of defaults? Can people just order whatever they want, and spin up dozens of VMs just because they feel like it? How do they pay for it? Is it just reported on, or does it link into internal charge-back or cost allocation processes? How?

I don’t want to be overly critical, because Platform9 only just came out of stealth. I’m not expecting the product to be as mature as ones that have been in the market as GA for over a decade, obviously.

But I am looking for a clear line between who their target customers are, what they value, and what Platform9 provides, both in terms of features and benefits.