Talk about a secret weapon. Altair will be among the first to offer high-performance computing (HPC) capabilities in the cloud, along with a ready-made simulation software suite, in part because of patented technology it built to give it an edge in software licensing.
One of the more notable but likely underrated features of Altair's HyperWorks simulation suite is a flexible licensing system that replaces the traditional and expensive one-user/one-license plans with a pay-per-use system. Organizations simply purchase a pool of what's called HyperWorks Units (HWUs), which are then applied based on what software is being accessed by how many users.
The token-based system automatically keeps track of who's using what, and once a job is finished, the tokens or points are added back to the pool where they can be applied in an ongoing, flexible basis. While Altair offers its full array of simulation tools in this fashion, the selection doesn't stop there. Through a partner alliance, a growing number of companies are also tapping the licensing technology and serving up their complementary tools in the same manner. To date, Altair officials say there are upwards of 30 titles now tuned to work with its licensing scheme.
So what does all of this have to do with HPC capabilities in the cloud? Well, to deliver on-demand HPC horsepower, there needs to be a couple of things in place, one being a flexible licensing and provisioning infrastructure (i.e., the licensing technology). There also needs to be a cloud datacenter, which Altair has built out and is now opening up access to with this new HyperWorks On-Demand (HWOD) offering.
The next question is why put HPC in the cloud? Altair says there's demand from companies that may have big simulation jobs that require the muscle of dozens, maybe hundreds, of multicore processors, but their requirements ebb and flow based on project status, so they don't want to invest in building out costly infrastructure.
"Say a company has a real spike in demand for compute infrastructure, but doesn't have the budget to buy another big Linux cluster -- more and more companies are attracted to cloud computing because of spikes in demand," said Martin Nichols, Altair's executive vice president, global alliances and operations, in an interview. Why turn to Altair? Here's Nichols's logic: "If they can access the same applications they're using in-house, but run them remotely and have them be managed by the company they buy the software from -- that should attract them to us."
Altair isn't the only company that sees potential in the HPC/cloud combination. IBM just this month announced an engineering-specific HPC cloud that emanated from an internal development effort around the POWER7 processor.
In addition to a scalable HPC infrastructure, HyperWorks On-Demand delivers Altair's suite of simulation software in a pay-as-you-go model. Currently available on HWOD is RADIOSS, a multi-disciplinary finite element solver for structural analysis; OptiStruct, a structural optimization solver; AcuSolve, a finite element-based computational fluid dynamics solver; and BatchMesher, a high-fidelity finite element mesher for large assemblies. Nichols says alliance partners' tools will also be available on the HWOD platform.
Here's how it works: Users first generate the model on their desktop, then upload their data to the Altair cloud to get the results. Sounds simple enough, but even Nichols admits there are some challenges. Running very large jobs with multiple terabytes of data may be problematic with today's bandwidth constraints. Also, with the compute resources being offered in a pay-as-you-go cloud model, that doesn't mean the process doesn't require a real understanding of the physics being modeled.
"There's a perception that if you put it on the cloud, it becomes easy," Nichols said. "But it doesn't really change the complexity of the problems you're solving or the set up required for some of the problems."