Cadence Chooses Platform9 For High Performance Computing Infrastructure Management

Semi-conductor tools company Cadence Design Systems has been using Platform9‘s cloud management solution for over 12 months and spoke to me about how it’s helped their research and development teams to make productive use of private and hybrid cloud.
Cadence bills itself as “the only company that provides the expertise and tools, IP, and hardware required for the entire electronics design chain, from chip design to chip packaging to boards and to systems.” People are learning quite a lot about the complex world of chip design and testing right at the moment, thanks to the news about Meltdown and Spectre issues with certain CPUs.

Cadence makes tools to help silicon companies design and test their products, which I learned is now vastly more complex than when I learned to do circuit modelling with SPICE as an engineering undergrad. Modern chips are multiple orders of magnitude more complex than the relatively primitive 68000 or 8080 series chips I learned on, and simulating them in software take some serious computing grunt to do.

This is where High Performance Computing, or HPC, comes in. Certain tasks in the design and testing toolchain benefit from massive parallelism, which needs lots and lots of CPU cores. However, these cores aren’t all needed by a single workload, or customer, all of the time.

Carl Siva, vice president of IT at Cadence
Carl Siva, Vice President of IT at Cadence

“We can over-provision cores on existing infrastructure, which is much cheaper than having dedicated infrastructure,” said Carl Siva, Vice President of IT at Cadence.

This is a classic capacity utilization problem that benefits from a portfolio approach at scale. Different workloads need access to resources at different times, so a platform that can be reconfigured dynamically can provide access to the same pool of cores to multiple workloads. It’s the same principle behind making your laptop look like it’s doing multiple things at once but is actually rapidly switching between different tasks, sharing the same physical resources (chiefly CPUs and RAM). Imagine having to buy a separate PC just to run Excel because your current one can only run Powerpoint.

But why not simply use public cloud? “We do both,” Siva said, “We proved the processes in public cloud, but there are some workloads that have to stay on private infrastructure.” Siva cited legal restrictions due to the contracted permissions relating to partner data, which many customers in multiple industries tell me is a major reason they need alternatives to public cloud.

“Private cloud gives us a lot of control and flexibility,” said Siva.