How a Service Provider can monetize OpenStack

It's been a while since the idea of sharing my experience with Service Providers struggling to find reasonable ways to monetize their OpenStack platforms crossed my mind, and I have now found the time to write it down. So I hope this post helps understand where the struggles are and ideally helps you get over them.

The first thing we need to understand is that the hosting industry, rapidly becoming the cloud hosting industry, is a highly distributed and rapidly growing one. It is very difficult to estimate the number of service providers, and I have found the best effort in Netcraft. Let's say there are several thousands service providers and a similar number of companies offering also web facing computers. What's important is the fact that there is a growing number of service providers and a growing number of hostnames, domains and active sites, which makes it an interesting business to invest in. That's why all telcos worldwide have their eyes on this growing industry, the hosting pie is getting bigger, and everyone has a chance of getting a piece of it.

So what are the challenges for OpenStack to become the platform of choice to this large set of service providers offering cloud services?:

Service definition:

Traditional hosting services (VPS, Shared Hosting, Dedicated Servers) are break even businesses. Whereas there are still opportunities for newcomers to make an impact, the case of Digital Ocean in the VPS (droplets, I mean) market is outstanding, these services have been offered for so long that the costs are already very efficient, and the investment in acquiring or retaining existing customers is growing. The trend is to compete in cost while leveraging some of the newest commodity technologies (SSD, better performance). Competing in cost is not appealing, and only the companies benefiting from scale economies (specially in HW) would be able to compete down the road in price. But that is not the smarter way to go.

It is leveraging new services and new ways to consume them what will allow companies to find their market, and that is where OpenStack, flexible and evolutive per nature, has its best chance to become the platform to orchestrate all these new services and models: hosted private cloud, virtual datacenter (isolated tenants or projects), different storage flavors, network virtualization, big data as a service, vdi, platform as a service capabilities, pay per use, prepaid resource allocation, hybrid cloud, federation, SLA measurement.

The challenge here resides on lowering the cost of integration and reliability, because not all of us are Amazon, Rackspace or HP Cloud and can benefit from their scale economies and in-house talent. For me, being an OpenStack native committed to its success, it all comes down to making the message clear to service providers and help them translate that message to their customers (who usually don't care that much about OpenStack, if at all). These services are already being delivered, it is not a promise but a reality. We just need to make them more available, simpler to understand, and billable.

Provision and billing:

The adoption of OpenStack is constrained by processes that are not part of OpenStack. I have pitched OpenStack to over 200 service providers worldwide, and the first question that invariably would come up is "How do I make money out of all this?". We no longer (I hope) are in the debate of Openstack being production ready, but we do face the challenge of making it enterprise consumable. This challenge was very well understood some companies. OpenStack is one of the main pieces of the puzzle a company is seeking to solve, there are several other tools and services required to make it all work for a specific company with their specific use cases (that will also impact the configuration and setup of the platform). A complete enterprise solution includes other tools around OpenStack and integration with existing processes. OpenStack is not, and can not be, an out of the box one-size fits all solution. It is a toolkit that enable companies to build their solutions.

For service providers, provisioning and billing are the most important ones because without them, the customer lifecycle is not complete. I do believe the success of Digital Ocean has more to do with these two processes rather than with the technology used. Well, that and a great knowledge database they put together, and probably a pile of marketing dollars helped along the way. Now, imagine if you could also have a neat provisioning portal and a reliable billing - not only metering, but billing (See notes) - automated on your OpenStack platform, and the ability to deliver innovative services and cloud consuming models, mentioned above, to your customers. Wouldn't that give you an edge? Would you be able to compete in other terms than just pure price?

Other key services like monitoring, application deployment, integration or migration procedures are either part of OpenStack or very well documented.

Deployment, HW, Support, Operation:

I have grouped all these 4 elements for a reason, they are answered right away. The question you need to ask, as a service provider, when considering using OpenStack is "Do I have the capability to deploy and support OpenStack based services?". The answer is probably "no, but I can do some of it". I believe, because it is in their DNA, service providers probably have the most cost efficient hardware purchasing procedures in the industry. Specially if they have been in business for years, they know their costs, they have a well stablished network and are not permeable to marketing. They are a nightmare for hardware salesmen!.

So the challenge they face has to do with deploying and supporting OpenStack and whether they have the talent or they need to hire or outsource or use one of the many distributions around. And once they choose, decide how they are going to operate and support their cloud. So here's the model I believe will suit the majority of service providers efficiently: buy your own hardware and operate your cloud (you might need to learn a few things, but overall, you've been doing this for years pretty much the same way, learn and apply the lesson). Look for assistance to deploy and support it, that is not your battle and can it usually brings frustration and a feeling of time wasted. It will give you peace of mind to focus on your business model and also reduce your time to market strategy, and that is extremely important.

Some final notes

• Time to market is important. Down the road, I foresee consolidation of service providers or acquisition by players (Telcos and other IT Companies), it is important not only to have an innovation edge (OpenStack is the best available driver because of its outstanding evolution), but to show it as billable services.
• The most important part of your cost analysis is still hardware, furthermore when you do not have to pay software licenses. Get it right by doing a proper sizing of your infrastructure, and that implies considering which services you are going to offer and how the different elements of those services relate (ratio RAM/vCPU/storage, performance vs density, etc)
• When deciding your platform, keep in mind there are two perspectives to consider: In the shortest term, you need reliability, cost and time to market strategies for your services. Down the road, you need to ensure the ability to deliver new services constantly, innovate and guarantee your are not locked in with a vendor (hardware or software). For both perspectives, OpenStack has the best answers.
• There are also lowest cost strategies for all those service providers with less than 20 servers that are working locally or in specific niche markets.
• Billing vs Metering. This would be material for yet another post, but essentially, metering allows you to know what's going on in your platform and billing allows you to create products, plans, define metrics and cross that product definition with information obtained from the metering system to produce a raw invoice.
• What if you changed billing for cost allocation? Wouldn't this very same analysis apply to all IT departments delivering their private cloud services to their internal customer? Just saying.