草榴社区

Why isn’t everybody doing it?

Run faster, run leaner, innovate sooner, save money, understand your engineering platform costs better and get best-in-class Business Continuity thrown in for free…  Why wouldn’t you want to move all your compute-heavy IC design engineering jobs to the cloud right now? Why aren’t you doing it?

Cloud offers a levelling-up opportunity for smaller organizations to compete against larger ones without the need for huge initial infrastructure investments. For larger operations cloud brings capacity flexibility when there are workload peaks and troughs to deal with. And for all hardware engineering teams, cloud delivers high levels of availability, reliability, scalability, performance, and affordability that are becoming harder to match with on-prem solutions – time to investigate further!

Why Cloud? Why Now?

IC design teams have resisted a move to cloud for hardware development based on several myths such as cloud’s perceived lack of security, a belief that it’s easier to predict availability of compute when you own it all on-prem, that cloud is not affordable, plus genuine questions about ease of use and how to move jobs to the cloud. These are all valid concerns, but as with most myths, better understanding can reveal a different picture. Read more about the myths.

Can you afford to ignore it…?

Today, there are 3 key drivers coming together in a perfect storm:

• The SysMoore era is driving systemic complexity and hyper-convergent design flows, which in turn require exponentially more compute and EDA resources.
• AI entering design tools and workflows is driving further multiplicative requirements for flexible (i.e., unlimited) access to compute and EDA resources.
• The good news is that CSPs have scaled HPC-optimized infrastructure, availability, affordability, and capacity to handle these scaled-out workloads.

Lifting and shifting workflows – talk to me about the challenges

Candidate workflows need to be encapsulated so they can be sent to the cloud with all the job dependencies packaged up with the job. For teams looking to port in-house workflows, this may mean some analysis of the existing on-prem workflows to establish the I/O requirements and file dependencies of each job, and then possibly some re-architecting of the workflows to make them cloud-ready.

Adoption models – how does it work contractually?

• The Bring your own Cloud (BYOC) model may be much more attractive if have already established cloud capabilities with a chosen CSP, and where demands can be very bursty and peaky. Licenses are not limited in this PAYG consumption model, similar to the CSP compute model. Metering and analytics are used to bill for usage and to allow for effective consumption budgeting. 
• With Software as a Service (SaaS) the user isn’t concerned with set up and running costs of using a public CSP as the tool vendor is hosting the application on the cloud service; think Office365.
• A Hybrid Cloud approach is the strategy for larger organizations with a pre-existing investment in on-prem capabilities. Peak demands can be met by bursting capacity into the cloud for suitable workloads, while less portable workloads continue to be run on-prem with no impact.

An opportunity to do things differently – find out more

Cloud is a major inflection point allowing engineers to be more innovative, to improve productivity, performance, turnaround time, and accuracy. Whether you are a small or medium sized organization with some or no on-prem investment, or a large established organization with significant sunk costs in infrastructure but need to be able to meet demand expansion requests or peak usage levels in a more sustainable way as the business grows, CSPs and EDA vendors now offer usage models that make cloud adoption affordable and scalable for IC hardware developers; Pay As You Grow with cloud!