SANTA CLARA, Calif. – Big EDA companies have their eyes on cloud computing, but their feet are still on the ground, according to a panel discussion at DesignCon here.
Cloud computing refers to software or services available over the Internet, but running on remote data centers typically managed by third parties. Some big consumer Web sites such as Netflix and business software providers such as Salesforce.com run as cloud services, but when it comes to design software—not so much.
"In the EDA industry we are still in a learning mode to see if it will work or not," said Hasmukh Ranjan, vice president of information technology at Synopsys.
Synopsys has used Amazon Web Services (AWS) for two years internally as a way to test its tools and the potential for cloud services, Ranjan said. But it is still gathering customer requirements for cloud computing and has yet to offer an external service, he said.
Cadence has taken a few maiden flights into cloud services. The company lets some customers use public cloud services such as AWS to try out its tools. It has also helped some customers install Cadence software on their private systems for sharing via the internal networks with the customer's branch offices around the world.
"For interactive EDA tools, the network bandwidth is not there yet," said A.J. Incorvaia, a vice president of R&D at Cadence. "But you can use remote clouds for storage or for batch services such as verification and simulation," he added.
Although two of the biggest EDA companies are not offering EDA software as a service, a few others are. MatLab's entire suite of software programs is optionally available on Amazon's EC2 cloud service and a handful of smaller companies are moving design and engineering software to the cloud as well, said Deepak Singh, business development manager for Amazon's EC2 service.
Licensing models are another hurdle. Some companies are using traditional licensing models on cloud-based delivery. But ultimately the software-as-a-service approach is what works best for cloud-based software, said Singh.
Meanwhile Amazon and others are catching up with the need for high-performance computing clusters on the Web. In July Amazon made cluster nodes available with 10 Gbit/s Ethernet links, initially showing off a 42 TeraFlop capability, Singh said. Since then it has also made general purpose GPU computing nodes available as well, he added.