HP's little learning dangerous in the cloud

The huge gulf in understanding between the traditional universe of enterprise computing and the new world of cloud computing was laid bare on stage at last week's HP Discover conference.

In what now appears to have been an ill-briefed gaffe, HP's chief operating officer Bill Veghte (pictured) announced on the closing day of the event that Workday had abandoned Amazon Web Services in favor of HP Cloud Services. According to CRN, which broke the story, the news was greeted by "thunderous applause from conference attendees," as Veghte explained:

"They are doing that because they want a cloud that they can rely on as an enterprise serving enterprises with some of the most rigorous SLAs in the business."

Except that Workday isn't abandoning AWS at all. Amazon's CTO Werner Vogels quickly tweeted that Workday remains "a happy AWS customer," a line subsequently reiterated by Workday itself in a terse email to GigaOm's Barb Darrow, who — despite describing it as "dicey" — had picked up the story on Sunday.

HP has since confirmed that Veghte had indeed overstepped the mark in his address. In an emailed statement to Darrow on Monday, the company put the record straight:

"During our HP Discover event, HP was very pleased to announce Workday as a customer for HP's public cloud. However, we misstated the impact of that announcement on Workday's relationship with Amazon Web Services. We apologize for the mistake."

So how could an experienced industry executive like Bill Veghte, with all of the weight of HP's marketing resources behind him, make such an embarrassing gaffe? ZDNet's Larry Dignan puts his finger on the crucial factor in his analysis of the "kerfuffle":

"Bottom line: Calling a company a cloud customer isn't nearly as big of a deal as noting an ERP deployment."

A little cloud learning

What likely happened in the corridors and hallways of HP headquarters was that people interpreted the news of the Workday win in terms of their own experience of traditional enterprise software sales. An on-premise ERP instance runs in a single location. With that knowledge as a frame of reference, someone hearing that Workday is adopting HP Cloud and has been using Amazon might well assume that the news meant Workday was moving its multi-tenant ERP instance from one location to another. As Alexander Pope once said, a little learning is a dangerous thing.

The huge divide between classic enterprise computing and truly cloud-native architectures was something I discussed here last month when writing about Netflix, HANA and the meaning of cloud. Netflix director of architecture Adrian Cockcroft uses the well-chosen term "forklifting" to describe the way that most enterprises think of moving computing to the cloud. This is what the vast majority of HP's customers are doing, and it's largely what the HP Cloud is designed to support. That's equally true of any other enterprise-targetted cloud — Microsoft Azure springs to mind.

In contrast, an advanced cloud-native architecture like Netflix is explicitly designed to run across several different cloud datacenters so that it is not dependent on any single physical location. "We can lose a third of our infrastructure without our customers noticing and calling customer services," says Cockcroft — a claim that Netflix deliberately puts to the test on a regular basis.

Under the covers, Workday is moving to an architecture that looks more and more like the Netflix model. Over several years, it has been evolving its architecture into a set of distributed services, initially by taking discrete functions such as payroll, analytics and integration out of the core object server and making them standalone services that are called as needed.

More recently, the vendor has begun porting a significant portion of the underlying database layer to Basho's cloud-native Riak distributed key/value store. David Clarke, VP technology services, told analysts at last November's Workday Rising conference that moving to the NoSQL database technology would allow Workday "to store many petabytes of data across multiple data centres."

This capability in Riak to run in more than one datacenter brings greater fault tolerance and elasticity into the core architecture, bringing it a lot closer to what Netflix is doing and even further away from the classic, tightly coupled, vertically integrated enterprise computing stack.

Workday keeps quiet

Workday hasn't divulged the precise details of where it runs its systems, and is saying even less than usual in the wake of Veghte's gaffe. With HP being one of its largest HRM customers, it is wisely hoping the fuss boils over without having to get drawn into the debate. I have had no reply to a request for comment I fired off by email earlier yesterday, a lack of response that doesn't surprise in the circumstances.

What is known is that Workday operates the bulk of its systems in colocation at Equinix, with a recovery facility at Savvis. It is using Amazon for some its more elastic processing needs. What it's putting — or thinking of putting — into the HP Cloud is likely to be a non-core service, while it beds down its use of the Basho Riak technology in the systems that it has under more direct control. But that is all informed speculation on my part rather than confirmed.

What is clear is that Workday's needs are very different from the "workload portability" HP is touting to traditional enterprise customers for its Converged Cloud offering. Moving existing computing workloads to a cloud host may have some benefits in terms of better economies, connectivity and elasticity, but it's no different in essence than the now-discredited ASP model, aptly characterized as SoSaaS — Same old Software as a Service.

While the old guard is finally moving to the cloud, vendors like Workday are staying ahead of the game by rearchitecting to become truly cloud-native.

Disclosure: Workday is a diginomica premium partner.

Photo credits: L plate © Brian Jackson - Fotolia.com; Bill Veghte headshot courtesy of HP