HPC YEARBOOK 2021/22


demanding HPC, storage, cloud and AI workloads. Best known for its work in higher education, the company has also delivered HPC solutions to many other industries, including engineering, life sciences, motorsport, public health, defence, and oil and gas. The company was historically a HPC


O


integrator, but has been increasingly moving into AI and machine learning as the workloads continue to increase and gain importance within the research. community. How has AI begun to impact your business and the HPC market? As you can imagine, over the last 10 or so years, our solutions have been used more and more for AI and machine learning workloads – inadvertently, in a way. I remember when the Nvidia Tesla GPUs first came out, when we were doing a 700-node supercomputer, and we’d have one machine in there with a couple of GPUs. Obviously, that that’s grown now, to a point where quite often the solutions that we’re supplying are 20 to 30 per cent accelerated computing. That is a combination of traditional workloads that are being accelerated with those GPUs, but also standalone, machine learning and AI workloads, as well. We are a HPC company that’s been


moving into AI. What we want to do is build on that expertise that we have, and try and take the skills of delivering these AI infrastructures and take those into this new market. HPC researchers are already adopting AI, but we are also speaking to all sorts of different organisations that are starting to look at these technologies. They are not necessarily going on Google and searching for HPC, integrator, even though they are doing what we would call HPC. We have a strategy to build on: the infrastructure skills that we already have, storage, networking, GPU, and other forms of accelerated computing, but then we want to build other technologies. On top of that, firstly, it’s some of


the different tools needed to help people manage how they use these workloads. We have also developed other partnerships that allow us to stop only talking to people about the infrastructure they’re using. We now have the capabilities to be able to offer that end-to-end solution, encouraging the infrastructure, but also the software


www.scientific-computing.com


CF is an experienced integrator that offers bespoke solutions and services to meet


and professional services to develop that solution for them.


Developing partnerships to help scientists better manage resources Coming from the HPC world, we’re used to the resources being managed by a scheduler. And we’ve got customers doing different things to allow them to run more AI workloads in there, where people generally want a dedicated machine to themselves for a period of time – especially if they’re developing these applications. So you’re not using a scheduler in the manner it’s intended, in some ways. And some of our customers are not using schedulers at all for these kinds of environments – essentially scheduling time on these machines using a Microsoft Teams calendar or something like that. That is really where Run:AI comes


in, because it’s designed for people to be using these resources, how the consumers of AI systems want to be using them. That’s a tool that helps you slice up machines as well. So you can have multiple users using the same physical piece of hardware, and potentially the same physical GPU as well. People want to have a resource provision for them, rather than them taking a workload and flinging it at a machine and expecting a response. When we talk about Run:AI, that’s the


hardware and infrastructure, and that’s more of a management tool. Whereas Clarifai, would sit above that, and could potentially be running on machines that are provisioned by Run:AI. Clarifai really is a piece of software or a number of different tools, essentially, that enable you to build your AI applications. So if you want to do the counting sheep in the field example, that is a piece of software you use to build a model and analyse that video or that image data. It comes with models already included as well to help people get a start. There are already cars and things like that in their database, for example. So you can start to build you’re own your own AI application, using that toolkit.


With Run:AI, anyone running these


kinds of workloads will get benefit from being able to better utilise the hardware that they’ve got, whereas Clarifai depends on what the challenge that you are trying to address, what problem you’re trying to solve, essentially. You’ve got some startups that are working on the medical side of things;


really focusing on one very specific application


“


really focusing on one very specific application. It could be a specific kind of medical image – whereas Clarifai is a much broader product and that allows you to develop something yourself. It depends on what your challenge is, and whether it’s something that there is already a company or a startup that exists that is already trying to solve that challenge for you, in which case, it might make sense to go to something that already exists. But if you want to go build something bespoke for yourself, it’s an excellent toolkit to get you started.


Making use of data Organisations are storing huge amounts of data. People think of big data as being a huge thing, you have to have petabytes, and petabytes for it to be called ‘big data’. But I don’t think that’s the case if you


have more data than is easy for you to manage, and to find the things that you want; then it becomes, essentially, a big data problem. I think, from a storage point of view,


there’s been a lot over the last couple of years around data management, and especially tools that are managing metadata as well. There are a lot of tools out there


that are now starting to people help understand what they’ve got, and help them take advantage of the value that, they’ve essentially been creating over the last X number of years. Software tools are helping scientists to find a way of taking advantage of that, using the data that they have been storing. So there are quite a few of these data management-type products that sell on top of storage – which, I think, can really help people find all the necessary images that they can then use with their AI programmes, for example. l


23


You’ve got some startups that are especially on the medical side of things;


“


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42