Engineering Computing News
Engineering Computing Resources
May 1, 2014
Getting into the cloud is simple. It takes you less than five minutes to sign up for an account on Amazon Web Services (AWS). Once within, you’ll be confronted with Amazon EC2 for general computing ($0.07 per hour and up); Amazon S3 for storage ($0.03 per TB per month and up); Amazon RDS for on-demand database; Amazon Kinesis for massive real-time data streaming; and much more.
But what is the sensible storage capacity and core count that would let you run your computational fluid dynamic (CFD) or rendering job? How do you apply the computing cores or the hours you’ve purchased to your turbine engine optimization session? Suppose you have chosen to migrate all construction documents to Amazon. How do you let the other 250 people involved in the project access the same documents? Who gets to edit them online? Who gets to delete files? How would the ones without AutoCAD installed on their machines look at the DWG schematics of a plant posted to Amazon?
“Some of our best customers are the ones who’ve tried [to implement cloud computing] themselves,” says Joris Poort, founder and CEO of Rescale. “They understood enough about the potential value to be willing to take the risk, but as soon as they delved into Amazon or Azure, they discovered it was a computer science problem, not an engineering problem.”
In 2011, Poort and Adam McKenzie, both former Boeing engineers, founded Rescale to offer on-demand simulation solutions by bundling hardware, software and middleware. But you won’t find an air-conditioned room with server racks in Rescale’s San Francisco headquarters. Rescale uses a global network of hardware infrastructure providers as the foundation for its own products. The company adds its own software on top of a computing hardware cloud to run and monitor simulation jobs. Should the user need additional licenses of simulation software to complete the task, the company also serves as the middleman to negotiate and acquire the necessary software. In essence, Rescale has repackaged cloud hardware infrastructure into a type of service—commonly known as Platform as a Service (PaaS)—for engineering firms that heavily rely on simulation.
The model points to one way engineers and manufacturers could get into the public cloud. Few design engineers have the inclination to figure out how to refashion the public cloud into deployable applications; they’ll most likely turn to cloud-powered solutions someone else has developed to tackle their problems.
Repackaging the Cloud
Barely in its third year, Rescale is still a new kid on the block, compared to established simulation software vendors like ANSYS, CD-adapco, COMSOL or Dassault Systemes’ SIMULIA. Even so, in its relatively young life, it has witnessed manufacturers’ changing attitude toward the cloud, Poort states.
“A lot of companies have decided they’re definitely going to do something,” he continues. “In our case, the high-end engineering customers realize they cannot just keep adding hardware to their infrastructure and remain competitive. So on-demand hardware’s value is easy for them to understand.”
As simulation jobs grow more complex, so does the need for greater computing power, achievable by investing in additional servers. Whereas large enterprises may have no qualms about the cost of additional hardware, midsize and smaller firms are much more judicious. It’s a need Rescale hopes to fill by offering not just access to the cloud, but cloud-powered storage and computing services tailor-made for simulation-heavy businesses.
Since it launched its business, Rescale has also been gradually expanding its software partner network. It now encompasses household names in simulation, like ANSYS, CD-adapco, MatLab, Convergent Science and MSC Software. For traditional simulation software vendors who’re not ready to offer their own on-demand PaaS or SaaS products, partnership with Rescale serves as the bridge to reach new users they could not reach on their own.
A $500 Answer to a $200,000 Question
Overland Conveyor Co. (OCC), based on Lakewood, CO, came up with a solver that simulates particle behavior in bulk handling systems. Its discrete element method (DEM) software “explicitly models the motion and mechanical interactions of each object in the physical problem … then provides a detailed description of the velocities, positions and the forces acting on each object at different points in time,” according to the company. It proves to be an ideal product for the mining business, for example, which requires routine transportation of materials in bulk using automated systems.
To be effective, OCC’s software has to run on high-performance computing (HPC) hardware. The investment may not be a significant barrier to well-funded mining firms, but could prove prohibitive for others. This presented a dilemma for OCC, which felt its solver could also be marketed to other businesses—for example, agricultural plants using automated systems to process peas.
Clint Hudson, manager of DEM applications at OCC, recalled how the company’s decision to integrate the cloud into its offerings came about. The catalyst was a hypothetical question: How can you reach companies that can’t invest $30,000 to solve a $200,000 problem? “What I want is to be able to give these companies the chance to spend about $400 to $500 to get the answer they need,” said Hudson. For OCC, the cloud was the answer.
“In the past, (companies) just bought the software, bought the hardware, and used them as often as they wanted to,” Hudson reasons. But what if “often” is just once or twice a year? Then the initial cost of hardware and software is no longer justifiable. So OCC repackaged its DEM technology as on-demand offerings, accessible for as little as $28. (Typical analysis sessions, Hudson estimates, would cost less than $500—well within the range affordable to the targeted companies.)
Last December, OCC launched the services as DEMcloud for Bulk Flow Analysis under the Applied DEM division. Users won’t need to purchase the software outright; instead, they can pay for usage. Nor do they need to invest in their own HPC hardware; access to OCC’s 32-core computing server is part of the on-demand product. In case you’re wondering where OCC’s computing infrastructure is located, the company uses AWS to scale as needed.
Some businesses fear they might miss the cloud gravy train if they don’t climb aboard quickly. Alex Brown, CEO of 10th Magnitude, has seen more than a few of them. “The issue is, they have not necessarily identified a winning business usage of the cloud,” he says. “When we start asking questions about where they see [the cloud] adding value to their business, sometimes they have a hard time answering.”
Recently, 10th Magnitude began working with a heavy equipment manufacturer to develop a mobile app for maintenance personnel working at customer sites. “The cloud and mobile technologies allow manufacturers to support their staff’s field activities in a way that in the past would have been cost-prohibitive,” Brown notes.
Built on Windows Azure, 10th Magnitude’s mobile app allows service technicians to remotely call up customer-specific installations of mining equipment right from the site. The pilot project uses Microsoft Surface tablets, but when the app rolls out, the technician should be able to use any type of connected device, running on Android, iOS or Windows operating systems, Brown explains.
“That’s the beauty of the cloud,” he adds. “All data exchanges are quite standardized, so from the device perspective, it doesn’t matter what device you use to consume the data.”
Because the mine could also be “in the middle of nowhere,” in Brown’s words, 10th Magnitude is planning to give the tablet users the option to preload all pertinent documents onto the device itself. That gives the technician a chance to troubleshoot other equipment or check on its health, even if the site offers no connection.
Let’s Talk Bandwidth
For simulation software powered by the cloud, the data connection still remains a critical factor. In automotive and aerospace, the 3D data that must be viewed, meshed and processed online is often measured in gigabytes.
“We knew from the very beginning that upload and download bandwidth is very important. So we use a custom Rescale JAVA uploader that uses multithreaded uploading,” explains Rescale’s Poort. “That fully saturates the bandwidth available.”
Rescale customers routinely upload 20- to 60GB models without hiccups, he points out: “It certainly isn’t going to take half a day.”
10th Magnitude’s Brown concurs that bandwidth issue was a well-founded concern—“about two to three years ago. The good news is, since then, some very interesting high-speed networking capabilities have come online for large industrial strength cloud providers.”
Brown points to the Express Route in Windows Azure, which allows a direct connection between the cloud and the end user. “Today we’re talking 1GB per sec rate; in the future, it’ll go up to 10GB per sec,” he predicts.
Let’s Talk Security
Ask cloud skeptics and advocates to create a short list of issues that affect adoption, and security is bound to find its way onto that list. “I would love to say security isn’t an issue, but it is—just a fact of life,” says Brown. “I’m of the opinion that, for 99% of the world out there, the cloud offers a more secure alternative to whatever people are doing (on premises).”
Especially when working with biomedical and high-tech clients, 10th Magnitude has to ensure its cloud solutions are compliant with both PCI and the U.S. Health Insurance Portability and Accountability Act (HIPAA)—“some of the toughest requirements for security,” according to Brown.
In addition, many customers’ attitude on security has changed, Rescale’s Poort points out: “They’ve gone from assuming (the cloud) is insecure so they shouldn’t do anything with it, to evaluating the security. When our customers look at our end-to-end encryption, (Service Organization Control) SOC 2 compliance, and security protocols, they often find that it’s a lot more secure than what they’re currently doing internally.”
OCC’s Hudson finds that “some companies are reluctant because they feel there’s a possibility their data might not be secure. Then there are those who won’t even look at the cloud because they’ve decided it’s not secure. They just have a policy against it.”
OCC’s software uses 256-bit encryption for every bit of data transferred to and from the cloud, Hudson reports. Besides, in mining and conveyor operations, the risk of IP theft doesn’t make a convincing argument, as many operations are well known and standardized. (It’s a different matter for an automaker, whose competitive advantage comes from the secret design of its next vehicle model.)
“So at least in some industries, I don’t think the fear of cloud is a legitimate fear,” Hudson says.
Public Cloud, Once Removed
For small and midsized engineering and design firms, the low-cost public cloud offers tantalizing possibilities. It’s a way for them to bypass the need to purchase and maintain hardware that they may not use perpetually. If their aim is to use the cloud for data storage and backup, the setup is relatively simple; however, if they plan to use the cloud as an on-demand computing resource, it’s not so straightforward. The emergence of companies like Rescale, OCC and 10th Magnitude indicates small and midsized engineering businesses will have an easier time reaping the benefit of the public cloud through third-party companies that have developed solutions tailor-made to tackle their problems.