What do “Shrek 2” and SAP have in common? A lot, according to Hewlett-Packard. The infrastructure hardware company announced Nov. 7 that it will collaborate with SAP on research aimed at improving the efficiency of HP systems running SAP software—or software components, as the case may be.
HP Labs developed technology to help DreamWorks, the studio that created the “Shrek” movies, scale its rendering capabilities by moving the necessary computing resources off-site to HP Labs. Rendering for “Shrek 2″—the process that converts animators computer-generated wire models into finished frames by adding details like color, light and texture—required a data center with 500 HP servers connected to the DreamWorks studio.
By putting the computing resources off-site and optimizing use, DreamWorks was able to draw on computing resources during crunch time and shutter the systems during off times.
The same concepts apply to SAP and enterprise software, according to John Manley, director of utility computing at HP Labs, in Bristol, England.
“With animation, the requirements are very similar to whats done in SAP,” said Manley. “An animator creates a model and needs to turn it into the final animated film. That process of turning it into a final animation of film is very, very highly computer-intensive, with lots of CPUs.”
Manley gives the example of a midmarket customer running a supply chain management application. “You dont want all sorts of machines tied up to the supply chain, when youre not using the applications. You want to go to a service provider that can quickly as possible build the right supply chain management application that links into your database, so those machines can be closed down when theyre not in use,” he said.
Manley said HPs goal with its Adaptive Enterprise model for computing—the notion behind Shreks rendering—is to get the concepts of ASP (aggregated service provider) right.
The general idea behind the ASP model—one that was propagated in the 1990s and, with little success in its own right, morphed into the idea of software as a service—is to share resources as efficiently as possible by letting another company run your software in their data centers. HP Labs too wants to share resources as efficiently as possible, according to Manley—not only on the hardware side, but on the software side as well.
Which is where SAP comes in.
But for the right ASP model, all the software and hardware have to be configured correctly for optimization. “And you have to put all the correctly configured software components onto the correctly configured machines,” said Manley. “That is a massive task. Its part of the machinery that lies between what were doing with SAP.”
The agreement between HP Labs and SAP Research really covers two areas of work: a broad umbrella agreement between both companies that includes research into areas like SOA (service-oriented architecture) semantics and RFID asset tracking technologies and business process analytics.
The second part of the agreement is the two groups launch project dubbed Adaptive SAP, which aims to take the next generation of SAPs Business Process Platform, or BPP, and show how users can deploy and adapt systems on an adaptive infrastructure, according to Manley.
SAPs BPP is the next generation of SAP applications—composable and configurable components that enable users to build business processes out of pieces of SAP software. BPP is part of SAPs ESA (Enterprise Services Architecture) and should be completed sometime next year.
HP and SAP are working together to optimize those components—either separately or as part of a composite application, and licensed by SAP or a service provider—on HPs hardware. The project is designed to combine technologies in model-based automation, virtualization and policy-based adoption, with compartmentalized deployment and execution environments.
“The Adaptive SAP project could help customers use hardware and software when and where they need it, rather than gearing systems to run full force at all times,” HP officials said in a statement.
The project is a lot harder than it sounds. To be able to optimize a process on a server, HP and SAP are developing a model for choreographing IT systems so that the hardware and software continually adapt to a companys varying workloads and changing process needs. So far, the teams have mapped out a half-dozen models, collectively called the Model Information Flow, that define the different phases of model, adapt and deploy information.
The first model is a generalized process model—which shows how one would model CRM (customer relationship management), for example.
The second is a custom process model that would take the first, general model and adapt it for a very specific use—Wells Fargo adopting CRM, for example—that brings in functional and nonfunctional aspects such as security, performance and reliability (where HP comes in).
The third model moves from the business process model to an abstract system model, which would detail the abstract systems needed to come to a process model, according to Manley. “Its interesting and quite complex,” he said. “You have to have a very intimate knowledge of the system youre talking about—in this case, SAP BPP.”
The grounded model is the process of taking the abstract system components and mapping to resource classes to match performance, security and the like. A bound model then looks at which machines are available and assigns a machine of a correct class to that system component. Finally, the deploy model takes the resulting information and deploys it, both at the software and hardware level.
“We see marrying these models really being at the heart of applying these [concepts],” said Manley. “The user would only be concerned with the first, general model—the rest would be applied at the software and hardware level.”
Part of that work also includes other development initiatives ongoing between SAP and HP, including SAPs BI Accelerator, an appliance written by SAP that brings in concepts of in-memory computing to radically speed up query efforts. While SAP is using in-memory in a business intelligence capacity now, eWEEK learned earlier this year that the company plans to go much further into the enterprise with its in-memory capability.
While the end goal for HP Labs is 100 percent automation, the Adaptive SAP project will likely pan out over the next three to eight years, according to Manley.
“Weve got the beginning parts of the system built. Weve done a lot of work in the models, so we understand the generic terms of the six different models,” said Manley. “Weve brought the beginnings of systems that will use virtualization to do things like live system migration of applications from one machine to another without any interruption. Were putting together all the pieces.”
Financial terms of the deal between HP and SAP were not disclosed.