Join David Linthicum for an in-depth discussion in this video High-performance computing, part of Cloud Architecture: Advanced Concepts.
- [Instructor] One of the great things about cloud computing is it allows you to get access to compute systems and storage systems that previously you couldn't afford. And one of the best examples of that is HPC, or high performance computing. So Google, Microsoft, AWS, basically most of the cloud providers out there offer versions of high performance computing. These are advanced systems typically used for research, number crunching, things like that. So in the past these supercomputers, which they are, you know, such as the Cray computers and other kinds of high performance computing systems were just out of reach for most small businesses.
So even though you're running a small research organization you could use a supercomputer very well for crunching your numbers, to do what researchers do, typically you couldn't afford it. So you're stuck with a PC or stuck with a smaller base server but now since it's on-demand, you can get a piece of a supercomputer and able to leverage exactly what you need and only use the amount of time that you need, it's very cool. So typically supercomputers and high performance computing based cloud on-demand kinds of systems are going to be leveraged primarily for research organizations, but this doesn't mean that businesses trying to do advanced analytics, which many businesses are doing these days, aren't going to lease the HPCs for that reason.
So, it is really kind of a growing area right now in research organizations who are leveraging HPC and typically this is going to be, ya know, not sensitive data. This is going to be massive amounts of data that they need to crunch the numbers on to figure out patterns, to determine results of various systems. And then the analytics side for the businesses, we may be looking at strategic things such as the ability to look at sales over the last 10 years in terms of trending analysis and regression analysis in terms of how things are going to grow in to the future.
And a traditional computer, ya know, may return the results of these questions back to them in a couple of days where a supercomputer can return it in a couple of hours and obviously there's a strategic advantage in that kind of a speed versus the latency that it's going to cost you if you're leveraging a supercomputer each and every time you go off and crunch those numbers. So any compute intensive task is going to be something that you want to consider HPE for and you also need to look at what each of the cloud providers, provide you in the way of high performance computing.
High performance computing becomes kind of a commodity now in terms of access by various organizations that may need it. And so you as an architect need to understand that, so you can offer them computing resources that allows 'em to punch way above their weight and we're able to leverage those resources at a fraction of the cost as it would take it if we actually bought a supercomputer system.
- Microservices and containers
- Complex, disturbed, serverless, and composite architectures
- DevOps integration
- High-performance solutions