Optimizing utilization of existing HPC simulation infrastructure


Speeding up a semiconductor design flow can be improved by taking three particular technology leaps with Altair:
These technology leaps allow chip design teams in semiconductor industry to rely on high throughput computing workflows which optimally leverage three classes of scarce resource:
The best and most efficient chip design teams invest effort and financial commitment in optimizing the allocation and productivity of these scarce resources.
Shrewd investments result in incremental wins in 5 key areas, all adding up to significant improvement in ROI for the chip design process and faster time to market:
Five key areas where ROI may improved:
Altair Accelerator’s scheduling speed is ten times that of the competitive solutions we have evaluated. Performant license-based scheduling enables our designers and engineers to explore more designs, faster, so we maintain an agile team and optimize utilization of expensive EDA licenses."
Altair’s license-first approach to scheduling enabled Annapurna Labs to enhance its resource management. It not only gave us more control over resource usage and cost, we dramatically improved productivity and time-to-product through the Continuous-Integration development flow."
Accelerator improved our time to market for the Matterhorn USB4 retimer. Being efficient in engineering keeps us on track with goals, tapeout dates, and milestones."
It is not uncommon for chip design companies to exclusively allocate compute resources to distinct teams, technology groups or projects. This could be done by physically separate compute hardware or job queues that have exclusive access to subsections of a larger cluster.
Though guaranteeing access to compute resources to certain groups of user, distinct compute resources create inefficiency in that they are rarely all fully utilized while other users may be starved of compute resources at high workload times.
Unifying all compute resources into a single compute farm, under control of advanced workload management software ensures the highest possible throughput of simulation workload for all user groups.
Altair® Accelerator™ is industry's fastest enterprise job scheduler. It was designed with high throughput computing in mind and is perfectly suited to optimizing utilization of existing simulation resources in the semiconductor industry.
Neither the scale of the compute resources or number of jobs in the queue is of concern to Accelerator. Clusters in excess of 50,000 cores can have more than 500,000 jobs in the queue, which is handled by Accelerator without any delay or inefficiency. Accelerator can do this because it is an event-based scheduler, executing jobs immediately when resources are available.
HPC throughput is often hampered by jobs waiting for EDA simulation licenses. These jobs were allocated resources in the system without checking that EDA simulation licenses are available. The resources are then held while the jobs wait for licenses, severely reducing cluster throughput. The problem can be addressed with home-grown front-end scripts but this solution is difficult to maintain and evolve over time.
Altair® Monitor™ tracks the availability of EDA simulation licenses in real time. This information can be used to turn these licenses into resources that are also managed by die workload management system. In this manner, Altair Accelerator becomes a license-aware scheduler, that ensures jobs are not stalled in the compute system for lack of simulation licenses. They will only run when the required licenses are available, thereby optimizing utilization of existing HPC simulation resources.
Managing job scripts, file transfers and command-line options to run jobs add unnecessary complexity to simulation workload requests. Altair® Access™ provides a GUI to users which guides them in selecting appropriate parameters for simulation requests.
Leading semiconductor design organizations use a combination of on-premise and cloud computing resources to deliver timely on cyclically varying workload requirements. Altair® NavOps®combine these disparate resources into a single, simple to use entity for end-users and sysadmins. Appropriate resources can simply be requested for workload and NavOps will secure the correct resources accordingly and transparently to the user, while retaining total control for sysadmins and business administrators.
Semiconductor designers, algorithm developers and software engineers all work with complicated simulation workflows, based on a multitude of dependencies. These flows depend on the availability of compute resources and limited software licenses. Understanding the flow itself and optimizing it for inherent parallelism while avoiding resource conflicts improve throughput and make users more efficient.
Altair® FlowTracer™ does exactly that. It is an advanced design flow development and execution platform that provides users with unique flow visualization and troubleshooting capabilities for greater productivity.
Job submission in semiconductor workloads typically ask the user for information on resources that would be required to run the job. The answers to these questions heavily influence throughput in the system. Jobs that request too much run time, RAM or unnecessarily large compute resources will wait in the workload management queues for too long and will waste resources when they eventually run.
Fortunately, semiconductor workloads are usually highly predictable given an understanding of typical requirements for particular workloads. AI/ML tools from Altair can learn the these requirements and can offer suggestions to end-users to make it easier for them to submit job requests. In several circumstances, AI/ML models can even override user requests with confidence, thereby reducing complexity of workload requests for end-users and improving accuracy in workload requests. In either scenario, the system will demonstrate higher throughput and efficiency.
Altair® AI Studio and Altair® AI Hub are only two examples of how Altair technology reduce job submission complexity for end-users in semiconductor design computing environments.
Cost control and allocation is critically important in environments where resources are scarce and/or costly. In semiconductor compute environments this includes compute time both on-site and possibly in the cloud and also utilization of expensive EDA simulation licences.
Assigning cost to projects, technology groups, or product teams can be continuously monitored by measurable metrics like CPU core-hours.
Costs and performance measures can be monitored in near real-time with Altair® Panopticon™ and reported accordingly. This process automates and simplifies project management for both end-users and system administrators.
Time to market in semiconductor development environments are critical. Slowing down design work during compute system down time is clearly detrimental to time to market and cannot be tolerated. Excellent support and continuous development of the software that governs the compute environment is key to the success of chip designers.
Job management systems are bread-and-butter for Altair with decades of experience in developing different workload managers and ancillary software. These workload managers are backed up by excellent support that will speed up problem solving in the development environments making the compute system more stable, reliable and predictable.
Customer references testify to their satisfaction with Altair's software solutions and technical partnership.
Inphi produces high-speed semiconductor components and optical subsystems that enable big data to move quickly throughout the globe, both inside data centers and between broadly distributed computing facilities. The team at Inphi evaluated several competing solutions before selecting Altair Accelerator, which stood out among the competition for superior performance and Altair’s reputation for excellent customer service.
Read Inphi storyAnnapurna Labs was established as a fabless chip start-up in 2011, focused on bringing innovation to the fast-growing cloud infrastructure. Annapurna Labs adopted Altair Accelerator and uses its Rapid Scaling feature to reduce cost by at least 50%. Additionally, they are seeing faster incremental development and continuous regression with tighter control on costs and benefits from a detailed view into resource usage by projects and users.
Read Annapurna storyKandou is a rapidly growing leader in fast, energy-efficient wired connectivity. Its disruptive technologies include chip products and IP solutions designed to accelerate the evolution of electronics. Faced with rapid growth, the team at Kandou needed to manage workloads and licensing for their expensive EDA tools. Kandou's team adopted Altair Accelerator for job scheduling and Altair Monitor for real-time license monitoring and management.
Read Kandou storyHero is an end-to-end solution designed specifically for emulation environments, bringing proven job scheduling techniques to the hardware emulation domain. Hero treats multiple emulators as shared resources — just like regular compute nodes — resulting in higher utilization, more simulation cycles, increased throughput, and ultimately, improved ROI. Users can address all aspects of emulation flow including design compilation, emulator selection, and software and regression testing for a streamlined process that makes working in an emulation environment easier than ever. Detailed reports with emulation-specific metrics enable optimized performance and informed planning.
Read more about Altair HeroDesigning and debugging modern integrated circuits (ICs) is highly complex, requiring companies to assemble large design and verification teams to identify and resolve issues. Altair® Silicon Debug Tools™ is a high-capacity, high-performing solution that reduces complexity for design teams and empowers users of all levels to rapidly debug IC designs, significantly shortening the development cycle. Altair’s IC debug capabilities handle an exceptional range of system, register-transfer level (RTL), gate-level, and SPICE-level design abstractions for modern system-on-chip (SoC) platforms, ensuring that semiconductor designs reach the manufacturing stage fault-free.
Read more about Altair Silicon Debug ToolsPSIM has been a leading power electronics and motor drive simulation and design software for over 25 years. With an intuitive, user-friendly interface and a robust simulation engine, PSIM is the all-in-one solution that meets users’ simulation and design needs. PSIM can handle quick power converter loss calculations, motor drive efficiency calculations, conducted EMI analysis, and analog/digital control. PSIM also offers automatic embedded code generation for rapid control prototyping. Additionally, with PSIM’s various Design Suites, users can design power supplies, electromagnetic interference (EMI) filters, and motor drive systems quickly and conveniently.
Read more about Altair PSIMAltair® Accelerator™ - Industry's Fastest Enterprise Job Scheduler
Accelerator is a high-throughput, enterprise-grade job scheduler designed to meet the complex demands of semiconductor and electronic design automation (EDA) and high-performance computing (HPC). It’s a highly adaptable solution capable of managing compute infrastructures from small, dedicated server farms to complex, distributed HPC and cluster environments.
Altair® Monitor™ - Real-time Software License Monitoring and Management
Monitor is a license tracking tool that lets users and administrators get real-time insight into software license availability, usage, job status, and more. Monitor uses current and historical data to help optimize license spending and utilization for organizations of any size — including large, multisite environments.
Altair® Access™ - HPC Job Submission Portal for Researchers and Engineers
Access provides a simple, powerful, and consistent interface for submitting and monitoring jobs on remote clusters, clouds, and other resources, allowing engineers and researchers to focus on core activities and spend less time learning how to run applications and moving data around.
Altair® NavOps® - Hybrid Cloud Scaling and Cost Management
NavOps maximizes business impact and enables global operations with optimized, cost-efficient high-performance computing (HPC) in the cloud. NavOps works with a broad range of cloud providers and workload managers, empowering organizations to prioritize critical workloads and control operational compute budget.
Altair® FlowTracer™ - Mission-critical Dependency Management
FlowTracer is an advanced design flow development and execution platform that provides users with unique flow visualization and troubleshooting capabilities for greater productivity. FlowTracer provides flow visualization, analyzes flows, and identifies inherent parallelism built into today’s complex flows, optimizing use of compute resources.
Altair® AI Studio - Data Science Design Software
AI Studio offers a visual drag-and-drop workflow designer for non-coding domain experts, automated machine learning (AutoML) tools, support for generative AI, and interactive data prep capabilities.
Altair® Panopticon™ - Comprehensive Data Visualization and Streaming Analytics
Panopticon lets business users, analysts, and engineers — the people closest to the action — build, modify, and deploy sophisticated data visualization and stream processing applications with a drag-and-drop interface. They can design visual user interfaces that give them the perspectives they need to make insightful, fully informed decisions based on massive amounts of data.