This article was originally published on my Cloud Power blog at Computerworld on November 10th, 2015
On October 21st, 2015, HP officially announced what many of us anticipated for months. After months of denials and flip-flopping they will shut down their HP Helion Public Cloud service. How does their SLA stack up against other cloud providers now ?
The problem with HP’s cloud strategy was simple:
In April 2015, Hilf told the New York Times. “We thought people would rent or buy computing from us. It turns out that it makes no sense for us to go head-to-head.”
HP offered a small subset of features cloud adopters look for. Their compute service offered a fraction of what AWS EC2 service does, at prices starting at 3-times those of EC2. Their relational database service never came out of public beta. Their storage services were pitiful compared to AWS storage options. They never offered any of the application-level services. Ultimately it was the lack of direction that undermined them:
It wasn’t a mistake to get into public cloud, but the failure came from keeping the product in beta so long and not being clear with a strategy that said why it was better, faster or cheaper than what was out there, Bartoletti said.
HP is a force to reckon with in the enterprise server market. For HP enterprise customers of today HP cloud may have made sense. It offered them a palatable transition to the cloud without giving up their investment into on-premise. Much like IBM in the 1980s and Microsoft in the 1990s, HP appealed to the C-suite. To paraphrase an old bit of wisdom, noone got fired from an enterprise IT department for picking HP.
Large enterprises of tomorrow are small start-ups of today. When they set out to build something they do not go to HP. They go to Google, AWS, or Azure. Each of these offer services that go beyond the basic compute and networking infrastructure. They do so in ways that appeal to application developers.
In a comment on my IT transformation post Twitter user
@ejohnfel said the following:
Second, certainly old-timey IT has to get used to the idea of moving the whole operation into the cloud, the reality is, the cloud only dislocates the physical hardware, the power and cooling requirements.
That is the type of thinking about cloud computing that HP counted on. Networking, servers, storage, and relational databases are the basic building blocks of application infrastructure. They are important but what makes or breaks the cloud is a universe of APIs and services for building applications. IBM, another big player in the enterprise IT market, knows this.
The attraction of the public cloud is in the following:
- Automation of all tasks and programmatic access to all capabilities of the platform
- Server-less compute model, i.e. AWS Lambda, Heroku, Google App Engine, and to a certain extent Docker
- Managed database products by use case: SQL, NoSQL, caching, analytics and warehousing
- Managed data processing services such as queuing, map-reduce, streaming, etc.
- Third party developer APIs such as Office 365, Salesforce, Google Apps, Evernote, Dropbox, etc.
- Open analytics APIs such as AWS Machine Learning, IBM Watson, and similar products by Google and Microsoft.
Application developers want to build applications. They do not want to manage infrastructure or take part in IT red tape. Where HP cloud could not succeed is at appealing to developers. They did not offer application services. They did not make it easy for citizen developers to build applications without IT bureaucracy.
HP cloud failure is a lesson to all trying to build IaaS, PaaS and SaaS clouds for their customers. Managed IT is not the future. Server-less apps, APIs and algorithms are.