Univa Solutions For Electronics Design Automation (EDA)

Top EDA Challenges

• Complexity and Scale of Environment

  With frequent additions of new applications, architectures, management tools and other systems, not to mention new users and their projects, the EDA IT infrastructure is a complex entity. In addition, the old days of tens of servers with a few sys admins to manually configure the environment are gone; now companies need to manage tens of thousands of servers… but budget for personnel hasn't increased accordingly.

  What sys admins need is for infrastructure configuration and provisioning changes to just happen – correctly, continuously and without manual intervention. In other words, they need the environment to be continuously and automatically optimized.

• Ever-Increasing Workload and Demand for Capacity

  Moore's Law is alive and well in EDA – the amount of workload that needs to pass through the environment doubles every 2 years, so companies need to get twice as much throughput every other year. Some EDA companies report over well 1 million pending jobs at a single site. And as workload size and complexity goes up, response times go down.

  This kind of demand requires new capacity from somewhere, but the solution of simply buying more hardware isn't scalable. Data centers are time-consuming and costly to build and maintain. At some point, scale-out ceases to be an easy or cost-effective option, so EDA companies must find ways of getting more efficient productivity from their existing compute resources.

• Underutilized Resources

  In EDA environments, underutilized resources are many: Multi-core systems not being fully leveraged…. Memory footprints not being fully used...costly large memory machines sitting idle waiting for large memory jobs to be submitted…. Parallel file servers, needed only at the end of workflow to satisfy large file I/O requirements…. and very expensive EDA licenses that aren't utilized 100% of the time.

  The bottom line is this: underutilized capacity equates to a higher cost per execution, and EDA companies need to find ways to optimize the use of these constrained resources.

• Time and Accuracy Pressures

  EDA is a competitive business, and time to market is everything. Tremendous effort is put into first time right silicon because the alternative is so expensive. Better to avoid these mistakes to begin with by providing engineers the capacity they need to identify errors prior to tape-out, because full mask re-spins will cost millions of dollars; even less expensive metal layer only re-spins still mean significant cost in dollars and schedule. Getting maximum utilization out of compute infrastructure and licenses means more simulations can be run and thus the better the chance to achieve first time right silicon.

A Proven Solution for EDA

Univa enables EDA organizations to create a more efficient environment where all resources are optimized.

Our UniCloud product was designed to support the unique requirements of an EDA environment, from providing a framework to optimize the utilization of computing infrastructure to the details of interacting with the EDA license daemons. In addition, our team – from executives to engineers – have strong backgrounds in the technology that powers many EDA environments. Combine this with our unparalleled ability to optimize along with our management capabilities for virtual and cloud environments, and you've got a solution vendor with no rival.

UniCloud provides transparent operation while retaining end user autonomy – so the user stays in control but is less involved with the underlying infrastructure. This enables engineers to focus on their core competencies rather than on IT. Univa also helps EDA organizations reduce the time lapse from design concept to delivery with increased computational capacity and reduced management complexity.

Key Solution Features

• Policy-based provisioning: UniCloud's policy engine applies rules to place jobs on the most appropriate infrastructure profile – bare metal or virtualized, internal or external – based on the requirements of the application or user.
• Checkpointing / job restart: Now engineers don't have to worry about outages or crashing ruining 2-3 week simulation runs. UniCloud can prioritize workload based on necessity (without residual impact) and create snapshots of long-running jobs to enable resurrection. With Univa, your job will finish in the timeframe you expected, period.
• Live migration: UniCloud can leverage scarce, idle resources (like large memory machines) while keeping that resource available for its intended function. UniCloud can dynamically move a job to a machine with more memory once the job starts, to access SWAP, and can load-balance workload after dispatch (so, no more "fire and forget") – all without impact to users.
  • An additional benefit here is in creating a "Green" strategy for the data center – UniCloud can move all jobs to a minimal number of machines to facilitate shutting down portions of a data center.
• Better dispatch characteristics: UniCloud offers faster dispatch of short jobs, with better fairshare, as part of optimization
• Support for virtualization: Univa can help you take advantage of virtualization – some customers have seen over 80% cost reduction when Univa products are coupled with virtualization implementations. Read the recent benchmark report »
• Cloud-readiness: UniCloud can add compute nodes by securely bursting to external computing environments on demand. UniCloud can also support multiple global data centers with transparent integration with external (or internal) virtual clusters. With Univa, when your business is ready for cloud, your technology will be, as well.

Unique Benefits for EDA

Optimize Utilization and Productivity

• Get more power out of existing hardware resources
• Support increased size and complexity of workload
• Avoid product spins by supplying the extra capacity needed to ensure accuracy

Maximize Application Performance

• Dynamically allocate applications based on resource availability, location and other parameters
• Continually monitor application characteristics and initiate infrastructure automatically, improving performance levels
• Enable checkpointing to minimize work loss

Simplify the Computing Environment

• Eliminate manual provisioning
• Enable IT to respond quickly to changing business needs that require adjustments to the infrastructure
• Streamline workflow to make better use of existing resources
• Abstract environment complexities away from the end user
• Provide the centralization, visibility and control to achieve a higher level of productivity

Reduce Infrastructure Total Cost of Ownership (TCO)

• Ensure maximum utilization of expensive resources
• Keep larger costs under control
• Free up budget to spend in more value-added areas of the design process
• Minimize application TCO as license utilization is optimized