Engineering Design Meets Cloud Technology
The way for engineering organizations to overcome numerous data issues is to transform siloed environments into shared engineering clouds.
Digital convergence has accelerated the breakneck pace of technology development and has put immense pressure on electronics manufacturers to develop new and innovative products even faster. All the sub-segments within the electronics industry — from consumer electronics to medical devices — are being forced to respond to lower consumer spending and stricter access to capital. In this environment the engineering community is being challenged to:
- Deliver innovative designs better, faster, and cheaper.
- Design high quality products with fewer designers or resources and across a distributed ecosystem of partners.
- Respond to increased CIO cost control of engineering IT.
Consequently engineering processes and infrastructure are being pushed to their limits, leading to increased demand on computing resources and rapidly escalating costs.
Additionally, the convergence of mechanical and electrical engineering and embedded software development is creating its own challenges. It would be fair to say that today, the design and engineering functions in companies are in a state of flux trying to grapple with these challenges.
Engineering environments have long been “protected” in the sense that their funding requests historically have not been subject to the stricter criteria of other corporate investments.
But as engineering organizations in the electronics industry, in particular have grown through acquisitions, they are being forced into an operational reality of siloed data centers tied to local projects and resources, limited or poor operational insight, underutilized resources lacking the ability to respond to peak demands, and designers who are tied to local desk-side workstations offering limited collaboration.
The way to overcome these issues is to transform siloed environments into shared engineering clouds — starting from a private and private-hosted environment and maturing to public over time. To achieve this, engineering functions require:
- Interactive and batch remote access.
- Shared and centralized engineering IT.
- An integrated business and technical environment.
This unlocks designer skills from a location, provides greater access to compute and storage resources, aligns resources to project priorities, and as a result companies realize improved operational efficiencies and competitive cost savings.
Cloud computing effectively reduces costs and complexity. It has the potential to drastically cut costs through improved asset utilization, and to simplify and reduce time to deployment across the enterprise. The ability to consolidate compute resources, virtualize them, and then dynamically provision resources from the common pool yields a much higher utilization rate and thus better economics.
In order to accommodate for peak demand periods, Electronic Design Automation (EDA) design environments typically do not operate at high utilization rates. On average they operate at around 50-60 percent, with peak utilization in the 90 percent range.
However, the cloud can help push the average utilization rates up to 80-90 percent while still accommodating for peak loading. The cloud concept is not alien — cloud based Web services including email, storage, and customer management software are already offered by Amazon.com, Google, Salesforce.com, and others.
Regarding security concerns, the reality is, even in today’s “cloudless” traditional business environment, there is a transfer of significant amounts of sensitive data:
- The fabless semiconductor industry is an example — companies today provide their design data electronically to the foundries, fully expecting the foundry to hold their end of the promise to maintain confidentiality and security.
- Managed services outsourcing companies deal with all kinds of sensitive data, from processing insurance to HR to financial outsourcing.
These concerns will be fully addressed as solution providers apply their core strengths in security and service reliability to cloud computing. For engineering, a well-considered cloud computing approach makes possible remote, collaborative product development and management between design, commercial, operations, and marketing teams. Everyone can access the most current information on demand, bolstering trust between the stakeholders and facilitating accurate and reliable decision-making.
Computing workloads can be broadly classified into two buckets — enterprise workloads and high performance, analytics intensive workloads. Forecasting application workload requirements accurately is a major challenge for IT. This challenge becomes even more acute as businesses increasingly depend on high-performance analysis and web applications, which have more variability than traditional enterprise business applications.
High Performance Computing (HPC) uses supercomputers and clustered computers to solve computational and data intensive problems. HPC helps enterprises achieve the speed, agility, insights, and sustained competitive advantage to deliver innovative products, increase revenues, and improve operational performance.
HPC cloud solutions are crucial in the semiconductor industry. The entire semiconductor design process — starting with product definition and design, Register Transfer Language design, functional verification, analysis, synthesis, simulation, and test are key HPC cloud candidates. The ultimate goal for many engineering R&D enterprises will be to virtualize the full semiconductor development process. Doing so will reduce cost by requiring fewer silicon experiments and improving time to market for next generation semiconductor technologies.
It will also drive the need for more HPC resources in support of predictive models for circuit design and process modeling. With engineering costs increasing exponentially in every dimension EDA ISVs that provide solutions used to collaboratively design, test, simulate, and verify integrated chips are being challenged to re-evaluate their business models.
For many years, IBM had deployed its own private HPC grid for developing future processors to meet stringent design schedules and quality goals on a budget for complex batch workloads in Logic Simulation, Design for Manufacturability (DFM), Layout versus Schematic (LVS), Design Rule Checking (DRC), Optical Proximity Correction (OPC), Timing, Synthesis & Build, etc.
Over the last twenty years, this has evolved into a true IBM HPC cloud engineering environment supporting over 3000 interactive Design Cloud users demanding rapid response on jobs related to Schematic and Logic Entry, Layout Entry, Design Auditing, etc. This solution is optimized for high utilization for a continuous stream of demanding batch jobs and enables remote high resolution graphics on laptops for interactive jobs.
One site holds design projects servers and data accessible globally by design engineers. Interactive jobs have highest priority to minimize designer wait time. The infrastructure also consists of a very large high performance cluster for OPC while large software and hardware co-simulations use a massive Field-programmable Gate Array system. The configuration can support many architectures and operating systems including IBM Power Systems and x86 clusters.
The IBM EDA Design Cloud Delivery Management System is optimized for engineering productivity — managing computationally demanding batch workloads with time-critical interactive user workloads. The IBM EDA cloud achieves 80-90 percent CPU utilization and 60 percent memory utilization with minimal overhead and no virtualization software to degrade performance and drive up costs.
Highly valuable designer time is also optimized for personal productivity. Stringent product design and development schedules and deadlines have been met consistently. In fact, IBM reduced the IT cost per designer by 50 percent and reduced their Power7 development cycle time by 25 percent.
IBM has now packaged and hardened the components of this internal HPC cloud as a set of new offerings to help clients tackle advanced scientific and technical computing workloads like analytics and simulations for product development. These new offerings include:
- IBM HPC Management Suite for Cloud: a new resource manager with tools to enable administrators to manage the cloud, schedule jobs and handle tasks like metering energy usage while allowing end users to view and access the full set of HPC resources available to them in the cloud.
- HPC Cloud Implementation Service: a new quick start service to help install, configure and optimize a private or private-hosted cloud with system administrator training to help ensure ongoing success.
- IBM Intelligent Cluster solutions: an integrated, optimized HPC cluster solution with servers, storage and switches — all factory integrated, tested and delivered ready to plug into the data center.
In conclusion, continued pressure on IT budgets will drive the need to improve utilization and return on compute dollars, and at the same time the reality of escalating energy and operating costs involved in building and maintaining data centers will force enterprises to adopt cloud computing models.
Other semiconductor companies have attacked the engineering asset utilization issue and have achieved limited success. Time to value and ability to achieve the full potential of savings continue to be challenges.
Companies will continue to investigate cloud solutions — private initially, transitioning to a hybrid cloud environment as tested capability grows - to help them manage development costs, while increasing engineering productivity and improving collaboration.
Given the complexity of engineering environments — mechanical, electrical and embedded software – and given the complex solution provider ecosystem and current licensing models, it will be a while before the utopian dream of a public integrated engineering cloud is realized. Until that shift occurs, we will continue to see pockets of private clouds integrated with key ISV software as the preferred solution.
About the author: Krish Dharma is IBM’s Electronics Industry Leader for the Americas. Mr. Dharma has a 27-year track record driving innovation and change within supply chain operations, information technology, and enterprise business transformation. He has held several senior level executive positions throughout the industry and has extensive experience within semiconductors, contract manufacturers, and consumer electronics. For more information on IBM’s Electronics Industry, visit: www.ibm.com/electronics.