HP's public cloud revamp to float down the road not taken
Analytics-as-a-service, Moonshot gear, and other 'crazy' ideas
Coming upgrades to HP's public cloud will see it pursue a different market from that dominated by Amazon, Google, and Microsoft – and that's a good thing for both punters and the company.
Like every other traditional IT seller, HP has been slow to wake to the rise of low-cost utility computing and storage, but on Friday the company told The Reg about some of the things it's preparing for its cloud.
In the short term, much of HP's cloud work is going to involve cementing the stability of its OpenStack-based platform and adding new features to give it parity with rivals. We'll be going into these announcements as they trickle out of HP Discover next week.
But there's stuff coming down the pipe that is far more interesting, and the reason why HP is able to do these things is because it got a bit of humility about its place in the market.
"Will [HP's public cloud] be as large as the McDonalds of the cloud? Probably not," HP's cloud chief Roger Levy told us. "There's room for a number of players depending on how you look at and segment the marketplace."
As we've said many times before, the amount of capital and intellectual investment made by Amazon, Google, and to a lesser extent Microsoft, in cloud technologies means all other providers exist in a different league.
But rather than try and duke it out in the land of the Amazon-wannabe clouds of Rackspace, Joyent, and IBM's just-acquired SoftLayer, HP is going for a different strategy.
Though compute cycles are far less variable than, say, a hamburger, there are different types of cloud slingers. Rather than going for stack-them-high and sell-them-cheap products, HP is going to focus on more expensive build-to-fit tech, such as analytics and integration that uses its own expertise.
"HP has a product called CSA – cloud services automation – which is a continuous-integration type environment. You might believe that we want to offer that in the public cloud," Levy told us.
"Just like in the analytics space we have Vertica and Autonomy – you might imagine those are things we're extremely interested in as a service," he said. "It's a logical set of assumptions that we'll be seeing some of those types of things."
Other cloud providers tend to offer either widely used tech, such as Microsoft via Windows Azure's SQL Database, or brand new systems such as Amazon's DynamoDB gear.
HP plans to walk somewhere between this by offering some of its own software remotely that has certain features that are hard to get in other products, without building completely new custom products. Autonomy's Bayesian inference techniques, though deployed elsewhere, have been refined within that product over years and years of sustained use, for example.
HP will also take a different path for its cloud infrastructure. As the world's number-one or number-two slinger of servers, it would be a PR nightmare for HP to go elsewhere for its IT gear. Other cloud providers are known to do this via manufacturing arrangements with companies such as Quanta or Wiwynn, moves that are likely behind the steady rise in the "other" category of servers shipped, as tracked by IDC.
Instead, HP is mulling the use of its massively "hyperscale" Moonshot server as the basis of its cloud.
"We have what's called a bill of materials evolution, in terms of as we order the next tranche of hardware, and each tranche gets better and better and better in terms of cost-performance, and Moonshot is lining up to be our next architectue," Levy said.
As with any speculation about, say, stock prices, we should stress that architectural decisions are prone to change, and HP could go for something else. But the use of Moonshot would provide the best possible environment for both stress-testing the server, along with case studies for potential cloud customers – cloud customers who are talking to HP right now about a "crazy" plan.
HP's "craziest idea"
There's one area where HP is going to differ with every other company out there: selling service providers complete cloud stacks.
Microsoft had planned to do this with Azure, but many of the promised Azure appliances never arrived, and when they did they were years late and uptake was muted. Cloud minnow OnApp has a similar scheme, but it pairs its software with appliance-style systems from Dell. IBM attempted to do so with CloudSystem, but given the Softlayer acquisition and the stated plan to move to OpenStack, a new strategy has not yet materialized.
HP, by contrast, is in the early stages of talking with service providers about equipping them with the gear and software necessary to run their own clouds.
Given the woeful state of integration within the OpenStack ecosystem, if HP could roll out its software across a broad swathe of telcos, it could assure data portability without needing to conduct potentially awkward negotiations with rival OpenStack providers such as IBM.
"It is a little crazy to take and enable potential competitors," Levy said. "Why in the world would you want to go and give away the secret sauce and build a public cloud for your competitors?"
"We think it's actually great for the industry. It's great for our customers. It goes through the scale question – scale can be achieved through a group of providers and not just the single provider. And at the end of the day it's a very good business for HP – it includes our software, our hardware."
If HP and the telco cloud customers stick to OpenStack for their clouds, and HP assures, as Levy puts it, total API compatibility, then data migration will be possible.
Though Amazon, Google, and Microsoft will likely remain kings of the major cloud workloads due to their experience operating vast infrastructures and their use of coherent mega-software stacks, HP could make a crust by offering certain premium services to enterprises.
If the company is really clever – and let us not forget that HP has a lineage of massive invention, save for the last decade or so in which it has been mired in political disasters and strategic missteps – then it has a chance to use its cloud as a way of testing and improving all of its products. [Excluding printers, surely?—Ed.]
Levy is certainly bullish on the idea:
Will [the HP public cloud] be significant? Yes. Will it have the ability to be a significant revenue contributor to HP? Yes. Will it be able to provide significant information back to the rest of the company to improve our offerings? Yes.
With much of the IT ecosystem fragmenting into separate specific companies with specific product lines, there are few integrated companies left that can make a stab at the cloud without being too cautious due to fears of cannibalization. We hope, for HP's sake, that it does the craziest things it can imagine. ®
Sponsored: Benefits from the lessons learned in HPC