Blog

The Future of HPC is Multi-Cloud

 

March 14 marks thirteen years since the launch of Amazon Web Services. One of the first iterations of modern cloud computing, AWS featured a modest three services: EC2 computing, S3 storage and a messaging queue called SQS. The goal of AWS at the time was to make it easy for developers to create simple prototypes and iterate quickly.

 

Today cloud computing has taken over the world of IT. AWS alone is a collection of hundreds of services, some of which could not even have been imagined in 2006. For example AWS Lambda, which lets you run your code completely server-less. With its blistering pace of product releases AWS has forced other cloud providers to keep pace, or die. Azure and Google are already carving their own niches, and IBM and Oracle are playing catchup.

 

Competition among vendors is a good thing. While it may seem wasteful compared to a top-down managed economy, it is competition that forces vendors to constantly innovate and stay ahead. You can thank competition for all those new features. When there's competition the customer wins.

 

But all this innovation and choice comes at a price. In the world of cloud HPC that price manifests as complexity and compatibility.

 

1. Complexity. Cloud vendors add new services every week. But customers don't have the time or expertise to change the way they work every time they want to use a new service. Customers need a relatively stable front-end that masks the complexity of the underlying, changing, morphing infrastructure.

 

2. Compatibility. Getting a complex HPC workload to run on one cloud is hard enough. But to take advantage of a multi-cloud world, you need to port that workload to each cloud. With multiple options for CAE software and cloud platforms the number of options explode. Customers need a solution that abstracts the different cloud without dumbing down features to least common denominator.

 

Enter HPC containers. Lightweight, portable, performant. Containers are the perfect antidote to the sprawl and complexity of a heterogenous world in HPC. Containers rely on the Linux kernel, so they can run anywhere there's Linux - which is everywhere. Containers are easily ported between clouds giving customer choice and flexibility. And containers provide performance that discerning HPC users need.

 

To learn more about how containers can help you win in the cloud game, give us a call.

 

Stay in the loop