For the second year now I’m virtually attending the DAC, the 47th annual EDA conference held this week in Anaheim. As a web oriented company we’ve yet to exhibit at the EDA industry’s biggest conference in Anaheim. Not being there physically, I enjoy all the info I can obtain remotely using twitter, blogs, and other online media. This year there seems to be more chatter about cloud computing and EDA - a topic that’s of particularly interest to a web oriented EDA company like PDTi.
Firstly, I saw a twitter post from James Colgan, the CEO of EDA community provider Xuropa, indicating that Kevin Bushby claimed that the Cloud is the only way EDA can grow. I’m assuming this is Kevin Busyby COO of FastScale Technology which was acquired by EMC, and who formerly worked at Cadence. While I agree that the cloud can help EDA grow, I’m curious to understand how Kevin and others see it growing.
Here are some ways I can see EDA growing using the cloud:
- Lower costs for compute resources could lead to larger EDA budgets.
- The cost and overhead of supporting customer on-site installations and evaluations could be reduced due the more controlled deployment environment of the cloud.
- The ability to use and get billed for tools at finer granularities could provide access to higher-end tools for companies that can’t afford the traditional EDA license models.
- The availability of limitless computing resources in the cloud could result in EDA users paying a premium to get that synthesis job or verification regression suite done more quickly.
- More visibility into how customers are using tools can provide opportunities to better server the customer, adding more value, resulting in greater revenues and profits.
- Hosting EDA tools in the cloud could eliminate piracy and add to revenues.
Some of these are things that we have already realized with our SpectaReg web application for register management and automation, which is offered onsite, hosted by the customer, or online, hosted by us. Whether hosted by the customer or us, the application is essentially the same, except the online user has the opportunity for some additional customizations. Interestingly, some of our customers are using virtualization technologies to create their own private cloud where they deploy SpectaReg onsite.
The great thing about the cloud is the ability to scale the compute resources, like RAM and CPUs on demand, and to have failover/redundancy available should some piece of hardware fail. If one has a fairly static requirement for these then cloud computing might not make sense. For example, a while back I ran the numbers on the cost of the equivalent of a dedicated machine would be on Amazon’s Elastic Compute Cloud (EC2). To have the equivalent compute resources available 24 x 7 x 365 via EC2 would cost more; however, a lot of machines are not used full-time and the compute requirements are bursty. This burstyness of compute requirements is where cloud computing really adds value.
To really take advantage of cloud computing, the application must be able to monitor/predict it’s load and be able to scale things up or down dynamically as needed. EDA applications or their wrapper scripts would need to get smarter to do this.
Another obstacle that critiques of EDA cloud computing often point out is that need to move files between tools and script flows. Technically, I don’t think this is an issue, aside from perhaps the need for EDA users to increase the bandwidths of their network pipes. Web service APIs would allow people to script all sorts of operations in their flows and move info between different EDA tools in the cloud, perhaps hosted by different cloud providers.
There are many other angles to cloud computing and EDA, and I could likely write 10 more blog postings on the topic. In terms of an end market, the cloud is a electronic system and there are opportunities for EDA to serve this growing market . Lori Kate Smith of ARM wrote up the 47DAC reception, mentioning how Mary Olsson of Gary Smith EDA cites Cloud computing as an application driver for EDA.
Another opportunity for EDA and FPGA vendors would be to have a cloud of FPGAs that could be re-configured. This re-configurable cloud computing would be pretty cool. Perhaps we’d need FPGA virtualization first, if it doesn’t already exist. Wonder if the folks at Google are looking into stuff like that…
Of course there is also the issue of security when it comes to cloud computing. I see Harry the ASIC guy was interviewing 2 cloud security experts at DAC and I’ve yet to check that out. Knowing Harry it will be worthwile. One concern is that if the cloud infrastructure becomes compromised then everything running on it can potentially become vulnerable. This is a bit different than a data center with isolated and distinct dedicated machines where each machine would need to be compromised individually.
Clearly there are a lot of opportunities and challenges for EDA with respect to cloud computing. It will be exciting to see how the future unravels. Stay tuned.