Arm increases operational efficiency through Agile methodology and cloud migration solution
MAKING LIFE BETTER IS NO EASY TASK
Arm makes technology that transforms the way people live. 70% of the world’s population uses Arm technology. With energy-efficient processor designs and software platforms enabling advanced computing in more than 230 billion chips, Arm must leverage the best tools, practices, and partnerships possible to guarantee the quality and timeliness of their IP (Intellectual Property) deliveries.
Due to growing complexity in IP design and verification, Arm’s on-premise fixed compute capacity lead to longer waiting times for resources, hindering their ability to run verification cycles in time, and increasing the time to market of their IP designs.
As part of an effort to improve the flexibility of their compute resources, and to reduce their global data centre footprint, Arm sought a partner who could help them move the majority of their EDA (Electronic Design Automation) workloads to the cloud.
A HOLISTIC APPROACH FOR A COMPLEX RESULT
Working closely with Arm, Endava first delivered a prototype to show it was possible to run EDA’s high throughput computing (HTC) workloads using spot capacity in AWS. This was then productionised into an API set, abstracting the underneath platforms such as Kubernetes across multiple clouds from the users, allowing them to run jobs in the cloud with portability.
With the prototype in place, Endava helped Arm develop a high-performance job execution engine utilising spot instance capacity in AWS. Through collaboration with AWS architecture and service teams, to tune the service and fit in feature requests, we delivered the capability to meet Arm’s scaling requirements. This facilitates the ability to auto-scale up and down rapidly to meet demand, consuming as much of the spare Amazon EC2 capacity as possible.
By migrating their EDA workloads to cloud, Arm overcame the limitations of on-premises compute capacity and gained elasticity through massively scalable computational power. The migration allowed Arm to reduce iteration time for semiconductor designs, and achieve the needed verification cycles without impacting delivery schedules.
Arm’s use of AWS Graviton2 instance types aids them in achieving high-performance and scalability, resulting in more cost-effective operations than by running on x86-based servers. Using instances based on Arm architecture for the design and verification of the new Arm IPs (Arm-on-Arm) demonstrates the benefit of Arm technology and drives the Arm ecosystem to have more EDA partners continue to port to the Arm architecture.
The job execution engine has already undergone several iterations, and the tech stack has included AWS Batch, API Gateway, Lambda Functions, Kubernetes, SQS, Kinesis, Rabbit MQ, RDS, DynamoDB, Event Bridge, CloudWatch, EFK, and tools including Terraform Cloud, Azure DevOps, Serverless Framework, Docker, Grafana, Influx DB, and Prometheus.
TEAMING UP FROM STEP ONE
To prepare for running EDA workloads in the cloud at large scale, Arm wanted their teams to be capable of developing microservices with continuous delivery and deployment, following the finer Agile methodologies. However, they weren’t just looking for consultants to assist them. Rather, they needed a partner to take the Agile journey alongside them.
Thanks to the recommendation of an Arm project lead with a prior, positive experience working with Envada, Arm became convinced we were the right partner for the job. Our reputation of teaming up with our clients instead of simply working for them preceded us, giving us the opportunity to help Arm see their Agile mission and cloud migration project through.
BETTER BUILT FOR A BRIGHTER FUTURE
Since their cloud migration, Arm has attained up to 6x improvement in throughput for certain types of workloads on AWS. Additionally, they have generated more engineering, business, and operational insights which assist in increasing workflow efficiency and in optimising resources and costs. Ultimately, migrating design and verification to AWS Graviton2-based instances reduced costs and enabled Arm’s flexible business strategy.
What’s more, the implementation of Endava’s solution helped give Arm the ability to scale to 350K concurrent vCPUs for simulation jobs, allowing Arm to meet customer deadlines and increase design standards. It has also allowed Arm’s engineering teams to run verifications without having to book slots weeks in advance.
Throughout our Agile journey with Arm, we have been privileged to get a clearer view of their long-term goals, giving us the ability to plan accordingly for any necessary team shape changes and profiles needed.