By: Ireland (boh.delete@this.outlook.ie), January 24, 2017 5:26 pm
Room: Moderated Discussions
RichardC (tich.delete@this.pobox.com) on January 24, 2017 5:08 pm wrote:
>
> Rendering high-quality 4K video is probably another reasonable app. Each frame might be
> 3840*2160*3*10bit ~ 29.7MB. You need enough DRAM on each node for the 3D model (games
> show that you can get fairly complex in 16GB). Pixar averages 3 hours/frame, but some
> > frames take 8 hours. At that average rate of 3 hours/frame, you need network bandwidth
> of 29.7MB/(3*3600) = 2884 bytes/sec.
>
> That doesn't count as a supercomputer app because it runs on a cluster of workstation-class
> machines. And I think that's part of what's going: you're thinking about the apps that run
> on supercomputer-class systems.
Richard,
As I remarked overhead, this is one of those situations as you describe, where someone is trying to create a new kind of solution, for a new kind of user (forget about thinking about it actually, in terms of whether the 'application' type exists, or doesn't exist already). The classic example in this one, is the transistor radio. The product called a radio already existed and it was technologically refined - but it sat in living rooms and was used to listen to classical music and radio theater - with a whole family gathered around it. The transistor radio wasn't different because it was a new application. It was different because the users who wanted to run that application, and the context in which they want to run it, was very different. I.e. Take it outdoors and walk around with the small radio device.
The challenge in all of this, is to try to imagine who could use existing HPC applications in today's world - but use them in a way that you or I - would never have thought to use them before.
This thing is going to be like the transistor radio. It's going to sprout legs and walk out the door, and do something that we haven't even thought of before. That's the idea of it. The product that the modern generation have rallied around, and it defines this generation in away that I find totally impossible to get my head around - is the water-proof, bluetooth 4.x, re-chargeable, wireless speaker - that connects to phones that people use to play music from. Like, why would people want to go swimming and rock climbing with their sound systems. When I was growing up, we just didn't do that. We broke back into vinyl records some time in the mid to late 1990's, and that was cool for a while. It still is (I think). However, we didn't trying to go rock climbing or swimming with our vinyl, that's for absolute sure. I did try to sit on a pile of vinyl records once, and the owner told me, how wrong a thing that was to do. I learned.
> And then you're saying because *those* apps need a lot of
> interconnect bandwidth, it's useless to build a system without a fast/expensive interconnect.
> But that's back to front: there are other apps that run on cluster-of-workstation systems
> (e.g. Beowulf clusters), and don't run on "supercomputers" precisely because the "supercomputer"
> system is expensive overkill for the app. But if you can build a
> flock-of-chickens or flock-of-turkeys system that actually gives more throughput-per-$
> and throughput-per-watt than a Beowulf/cluster-of-workstations, then it can be quite useful,
> even if it isn't a "supercomputer".
>
>
>
> Rendering high-quality 4K video is probably another reasonable app. Each frame might be
> 3840*2160*3*10bit ~ 29.7MB. You need enough DRAM on each node for the 3D model (games
> show that you can get fairly complex in 16GB). Pixar averages 3 hours/frame, but some
> > frames take 8 hours. At that average rate of 3 hours/frame, you need network bandwidth
> of 29.7MB/(3*3600) = 2884 bytes/sec.
>
> That doesn't count as a supercomputer app because it runs on a cluster of workstation-class
> machines. And I think that's part of what's going: you're thinking about the apps that run
> on supercomputer-class systems.
Richard,
As I remarked overhead, this is one of those situations as you describe, where someone is trying to create a new kind of solution, for a new kind of user (forget about thinking about it actually, in terms of whether the 'application' type exists, or doesn't exist already). The classic example in this one, is the transistor radio. The product called a radio already existed and it was technologically refined - but it sat in living rooms and was used to listen to classical music and radio theater - with a whole family gathered around it. The transistor radio wasn't different because it was a new application. It was different because the users who wanted to run that application, and the context in which they want to run it, was very different. I.e. Take it outdoors and walk around with the small radio device.
The challenge in all of this, is to try to imagine who could use existing HPC applications in today's world - but use them in a way that you or I - would never have thought to use them before.
This thing is going to be like the transistor radio. It's going to sprout legs and walk out the door, and do something that we haven't even thought of before. That's the idea of it. The product that the modern generation have rallied around, and it defines this generation in away that I find totally impossible to get my head around - is the water-proof, bluetooth 4.x, re-chargeable, wireless speaker - that connects to phones that people use to play music from. Like, why would people want to go swimming and rock climbing with their sound systems. When I was growing up, we just didn't do that. We broke back into vinyl records some time in the mid to late 1990's, and that was cool for a while. It still is (I think). However, we didn't trying to go rock climbing or swimming with our vinyl, that's for absolute sure. I did try to sit on a pile of vinyl records once, and the owner told me, how wrong a thing that was to do. I learned.
> And then you're saying because *those* apps need a lot of
> interconnect bandwidth, it's useless to build a system without a fast/expensive interconnect.
> But that's back to front: there are other apps that run on cluster-of-workstation systems
> (e.g. Beowulf clusters), and don't run on "supercomputers" precisely because the "supercomputer"
> system is expensive overkill for the app. But if you can build a
> flock-of-chickens or flock-of-turkeys system that actually gives more throughput-per-$
> and throughput-per-watt than a Beowulf/cluster-of-workstations, then it can be quite useful,
> even if it isn't a "supercomputer".
>
>