Transforming HPC Efficiency: AMS OSRAM's Journey to Scalable Solutions
Overview
In the ever-evolving landscape of technology, companies are continually seeking ways to enhance their operational efficiency and reduce costs. A leading global provider of innovative light and sensor solutions recognized the need to transform its high-performance computing (HPC) environment to tackle several challenges, including high total cost of ownership (TCO), performance bottlenecks and a lack of automation. With a rich history spanning over a century, the organization aimed to improve efficiency, scalability and visibility within its operations. By conducting a thorough assessment and implementing automated solutions alongside elastic infrastructure, the company successfully achieved significant performance improvements and operational efficiency. This transformation ultimately allowed them to align their HPC capabilities with the demands of modern, resource-intensive EDA simulations.
The Challenges
The customer is a global leader in innovative light and sensor solutions, with over a century of industry experience. It combines excellence in engineering and advanced manufacturing capabilities to create cutting-edge technologies that enhance safety, intelligence and sustainability. The customer faced a high TCO and lack of performance, as well as a lack of scalability and bottlenecks. The environment was also not automated and lacked visibility.
- The customer was facing a very high TCO with unsatisfactory performance
- The customer faced challenges in meeting resource requirements for running EDA simulations within their HPC landscape
- Lack of elastic and scalable infrastructure created bottlenecks and restricted performance
- Lack of a flexible, pay-as-you-go consumption model
- Lack of faster provisioning (infra and services) in response to business needs.
- Limited infrastructure visibility
- Lack of cluster automation and auto-scaling
The Objectives
To make the HPC landscape more efficient, enhance performance and remove bottlenecks, the TCO needs to be improved and a flexible commercial model introduced.
- Lower TCO and a flexible commercial model
- Improved performance and scalability
- Improved landscape visibility
- Introduction of automation for faster provisioning
The Solution
It helped the customer achieve their business goals by providing flexible commercial models, automating the HPC landscape and building a scalable HPC landscape.
Assessment
- Conducted a thorough assessment of the current HPC landscape and identified specific bottlenecks
- Identified the specific requirements and constraints of the client's HPC environment
Build
- Automated provisioning of HPC infrastructure using Terraform
- Automated provisioning of HPC workload Scheduler
- Auto-scaling of HPC nodes based on the EDA simulation
- Seamless user experience for running simulations in the cloud
- Data migration from on-prem to AWS
- Elastic infrastructure is able to meet the high resource requirements
Operate
- Seamless user experience for running simulations in the cloud
- End to end day to day operations management
- Continuous monitoring of HPC landscape
The Impact
The customer achieved enhanced performance, with a 30% to 40% performance improvement from on-prem and greater landscape visibility. Furthermore, there was a significant reduction in the manual efforts for managing the HPC environment.
- Enhanced performance with 30 to 40% performance improvement from on-prem
- Greater landscape visibility
- Lower efforts due to the implementation of automation
- Faster job execution time for all the small, medium and large simulations
- Zero cost of compute nodes if no HPC jobs are running
- Able to meet the high resource demands for running EDA jobs without any CapEx
- Zero waiting time for execution of the simulations
AWS services
- Amazon EC2 with auto-scaling
- Amazon Elastic Block Store
- Amazon FSx for NetApp
- AWS VPN
