Premise: Rolling your own private cloud using OpenStack has been a high risk strategy for the vast majority of enterprises. The first five years of OpenStack’s life, the first epoch, was characterised by over-marketing and under-execution, founders going out of business (Nebula Inc), diversifying away (Rackspace), losing talent (PistonCloud’s CTO moving to Pivotal) or being acquired by the five tech titans (HP, Cisco, EMC, IBM, RedHat). This OpenStack market correction has led to a change in the landscape.
In this second epoch it is expected that the OpenStack enterprise opportunity will divide into two sets: one is limited to the enterprise outliers (Paypal, Walmart, CERN, large banks), the second is the rest of the enterprises who don’t build but instead share the responsibility with the reduced field of providers and the coming federated titan clouds from HP, Cisco, EMC, IBM and RedHat. We will also see the emergence of higher-plane cloud solutions from Pivotal and others and a changed focus away from infrastructure and up to the application and data, relegating OpenStack to a lower, subsumed level.
Contents
- 2010-2014: The 1st OpenStack Epoch
- The Seven Sins of OpenStack
- 2015 and Beyond: The 2nd OpenStack Epoch
- Tech Titans Wrestle the OpenStack Baton
- The Dawn of the Federated Titan Cloud
- Final Thoughts
2010-2014: The 1st OpenStack Epoch
In 2013, according to an official Office of Inspector General Report IG-13-021, the US government had a change of heart about using Nebula in their “cloud first” strategy:
..after investing $19.7 million, NASA suspended funding for Nebula in 2012 and shifted its cloud strategy to the purchase of cloud services from public providers”.
The NASA Nebula project was the source for the Nova part of OpenStack and the then NASA CTO started Nebula Inc after his NASA days. Nebula inc carried on but its five-year lifetime ended on Apr 1, 2015, the market for private-cloud or service-provider-cloud powered by OpenStack seemingly running out of fuel. Was the feds reason for dumping Nebula a sign of things to come?
..tests found that the public clouds had matured to be more reliable and cost effective and offered much greater capacity and better IT support than Nebula.”
In other words, three years ago public cloud was cheaper and better than the internal Nebula project (precursor to the Nova part of OpenStack) private cloud…but the desire for private cloud is noted, along with the associated costs:
While the private cloud alternative enables agencies to manage their critical IT services and control access to sensitive data directly, these benefits come at the high cost of owning and operating data centers.”
OpenStack and Nebula Inc founder, Chris Kemp, stated in an interview in late 2014 that “NASA has a lot in common with large private enterprises”, so is it fair to expect those enterprises to have the same attitude as NASA: public cloud is better?
Enterprises would be startled to hear Chris say the following which rings alarm bells to the enterprise bearing the cost and the risk:
“One of the reasons that OpenStack is making consulting companies a fortune is that every cloud is a custom build… which changes with every release of OpenStack every six months.”
Andrew Clay Shafer, who was VP at CloudScaling and also at Rackspace, has a robust insiders view of OpenStack that all enterprises should read.
The Seven Sins of OpenStack
The seven sins of OpenStack that emerged in this first epoch must be addressed in the second:
| Commercials | Business case on buy and sell side: are enterprises getting value for money (see NASA), and how do vendors make enough money fast enough to sate investors (see Nebula) |
| Incompatibility | Lack of AWS-compatible API breaks the OpenStack-AWS hybrid cloud story, one that Eucalyptus – now HP – alone tried to fix). Integrations between OpenStack components has been inconsistent, and integrations between OpenStack and external components such as switches has also been problematic. Integration problems are hard but improvements are happening with the EC2 “drop in API” and a much bigger list of infrastructure integrations. This weakness could soon be a strength for OpenStack. |
| Difficulty | OpenStack is notoriously complex and difficult to implement requiring significant engineering ability, one of the reasons it might be ok for outliers like PayPal and Walmart with large engineering capabilities but not for the majority of the Fortune 1000. It takes significant talent to design, deploy and operate a custom OpenStack cloud and this can be a serious headache for the enterprise. It is normal to have to hire developers for each OpenStack component deployed, which is unusual for enterprises. |
| Rhetoric | The “OSS is best” and “Vendor Lock-in” shrieking in the industry along with OpenStack over-marketing vs. working features can deafen people to the real information and has lead the OpenStack players to be accused of the ex-VP of Cloudscaling of prioritising marketing over substance. |
| Focus | Should enterprises focus on becoming an IaaS provider, or a consumer of such a service? Enterprises should be really cautious and business led, not technology or industry-fad led, or repeat undifferentiated heavy lifting. |
| Delivery | The cost of the datacenter (don’t build a new one!), the infrastructure, the team to build it, the consultants and everything else: the IT industry is notorious for project overruns and enterprises look to mitigate these risks by purchasing a ready-made solution from providers with SLAs (and a big throat to choke, more on that later). |
| Operations | Not only are there brand new technologies to learn how to operate but the interface to consumers is also different. Enterprise operations and service delivery will have to adopt, this incurs a cost of change and risk. Don’t overestimate the desire for consistency, predictability and reliability with an operations group that isn’t rewarded for innovation but is punished for failures. |
If you are an enterprise and you are considering building or buying OpenStack then I encourage you to also read an account of how hard it is to “do” OpenStack. This article from David Laube, VP Platforms at Packet describes how they wasted four months of work.
Given the experiences of the last five years it is fair to deduce that OpenStack is not for the majority of enterprises to build or buy. It is still valid as part of the enterprise story if it is delivered by focused service providers who can mitigate the complexity and cost problems that enterprises don’t want to deal with, in fact given the reduced number of players in the market and the specialist capabilities required then this is an opportunity for providers.
Watch Stu Miniman of Wikibon featured on this videocast with Niki Acosta and Jeff Dickey:
http://youtu.be/76XCnX8mH9w
2015 and Beyond: The 2nd OpenStack Epoch
If you look at the market leaders / noise-makers in OpenStack, here’s what happened to some of the OpenStack players:
| 1st Epoch player | 2nd Epoch player |
| Nebula | <gone> |
| Rackspace | Reduced / diversified |
| Eucalyptus | Acquired by HP |
| CloudScaling | Acquired by EMC |
| MetaCloud | Acquired by Cisco |
| eNovance | Acquired by RedHat |
| Mirantis | On-demand OpenStack |
| Piston Cloud | Lost CTO to Pivotal in 2014 |
| Platform9 | OpenStack “as a Service” |
| BlueBoxCloud | Hosted OpenStack-powered Private Cloud” |
| IBM | Injecting OpenStack into Softlayer |
It remains to be seen what how the big five titans will succeed with their OpenStack approaches but these competitive giants are introducing fragmentation with multiple flavours of OpenStack distributions.
In the second epoch, the smaller pure-play providers (Mirantis, Piston Cloud, Platform9) are adding on-demand hosted models to their traditional consulting “pay us to help you build an on-premises solution”.
As an example, the ex-VMware veterans who have started Platform9 have an interesting “OpenStack-as-a-Service” SaaS model, where they run the OpenStack bits (and own the SLA and all the pain) and provide an interface to help you connect your on-premises infrastructure (network, storage, hosts) to the hosted OpenStack control plane.
For the tech titans, enterprises and cloud service providers are watching their acquisitions and strategic moves to federate cloud service providers around their technology.
Tech Titans Wrestle the OpenStack Baton
The majority of the leading OpenStack players have been acquired by the tech industry titans, and these titans of IT have good reasons to add OpenStack to their armoury:
| Talent | OpenStack requires talent and talent is scarce. It is a common business practice to speed up entry into a market by acquiring a product and talent rather than growing your own organically |
| Reach | The tech titans have enormous, global enterprise reach into many markets and regions. |
| Relevance | By glueing OpenStack to their product set, they keep their products relevant. |
| Vertical Integration | It’s a kind of vertical integration-by-substitution and removes the dependency on a previous partner-now-competitor’s virtualization and cloud stack. |
| Customer Service | The tech titans have giant throats to choke, and enterprises like to have that option for when problems happen. When an enterprise decides to buy, it needs to know that a CEO to CEO conversation can happen when the proverbial hits the fan. |
| Competition | If executed correctly, it is possible to build federated clouds of OpenStack service providers to offer an alternative to public clouds. |
This last point is pivotal to the bigger war picture between public cloud providers and the tech titans. The federated cloud model underway from titans HP and Cisco should be interesting to enterprises and service providers alike.
The Dawn of the Federated Titan Cloud
The tech titans are used to serving enterprises via a distribution and reseller channel and they are seeking to replicate this model in the cloud by building federated clouds of selected cloud service providers.
- HP’s Cloud28+ EU-only program could benefit service providers in a fragmented and disharmonious, twenty-eight state market by signing up to a central market place that will ease sales and fulfilment. If this marketplace works, then it will benefit enterprises of all sizes to find suitable services, do simple comparisons and grow into new regions. The benefit for HP is that the early adopter service providers will be customers of HP.
- Cisco’s Intercloud strategy is different to HPs also aims to deliver on the hybrid cloud approach and unify service providers.
This will cause some shake up in the cloud service provider market as allegiances are forged. It is likely that a CSP is not aligned to multiple titan clouds because the titans will be copying the disti/reseller model of creating allegiance via business benefits to the provider.
Final Thoughts
Due to the sins of OpenStack identified in this article, and the continual changes to the providers and products, deploying and operating OpenStack clouds should be left to the outliers, expert/niche players and titans/service providers.
The risk of “rolling your own OpenStack” is not just related to that infrastructure layer. Because IaaS is a foundational cloud layer then it’s risk to the business is measured by the probability of the failure of that layer, and the probability of it causing failures to the above workloads, multiplied by the impact of the failure of those workloads. Couple this with the cost and complexity and it is unappealing for enterprises to take on their own OpenStack unless they can find ways to mitigate them, such as partnering with an expert provider.
Action Item: Enterprises should avoid building their own OpenStack solution unless they are an outlier (massive enterprise with huge engineering capability). If the enterprise decision process leads them to choosing the features and functions provided by OpenStack then the best approach is to mitigate the risk and rent a private cloud from an expert provider which, in the second epoch, will likely be the new shared-responsibility/SaaS model like Platform9 or a service provider backed by a familiar vendor such as HP, Cisco, EMC, IBM and RedHat.
