I’ve worked on several systems that had enough flop/s to make it in the top 5-10, but for which we never submitted benchmarks. Sometimes their backend network layout technically would make them several smaller clusters for an HPL run, sometimes it’s because the cluster is too heterogeneous to get a good benchmark result, and sometimes it’s because the employer wants to keep a low profile.
Most of the time, it just that it’s a hassle. It takes a while to prep and tune a big hero run for benchmarking, and if you spend a billion dollars on a cluster, it’s making you a lot more than that. Taking it down for a day or two stops the money printers.
You still need to traverse physical segments in the wireless path: think receiving dish to the next transmitting dish, the end of the path to get from the trading systems onto the roof and into to the first dish, etc. Every nanosecond counts.
> I'm sure some HFT traders are going to make an absolute fortune, but I doubt it'll have a huge impact for most other people.
They’ve been using hollow core fiber (and funding research into it) for nearly a decade. I know it goes back further than the 2017 spinoff mentioned in the article, but https://optics.org/news/11/9/52 talks about it a bit.
> But what about Jump Capital? They have a crypto division that also does market making. The difference here is there are given a large chunk of tokens to market make with, as they please.
It may be a little early to make that comparison. Jump is still being investigated for its crypto shenanigans.
I think it’s a way of keeping the museum’s single most popular piece of art on display whilst working it. I think most museums would remove it for a while, but so many people come specifically to see this painting that they want to keep it viewable, so they make a little show of its restoration.
I dunno; I’ve been through that floor 5 or 6 times since they started work, and people always seem to love the spectacle of it.
Or have them set up in a way that makes them hard to run full-system benchmarks on. I can think of a couple of financial firms that have clusters that would rate on the Top 10, but a) as you suspected, they’re too busy running the money printers to take them down for a few days to run benchmarks, and b) they’re split up into more manageable little clusters, so the high speed fabrics don’t connect and allow every node to talk to every other node, which you need for an HPL run.
Do you have issues with other rollerball pens? I’m also left-handed, and really struggle with them (especially the G2). My issues seem to be that when I’m pushing the point across the paper, rather than pulling it across like a right-hander would do, they often dry up and there’s no ink coming out.
I like how quick drying gel pens are, but I haven’t been able to write with them consistently. If that happens to you with other rollerballs, I’ll give that Pentel Energel a try.
> In your experience, has this led to more careful hiring?
Not in my (albeit limited) experience in the Netherlands. What I’ve seen it lead to is people hired initially into temporary one-year contracts, which are then converted to permanent contracts if things work out. I’ve definitely seen people be fired or otherwise pushed out even on permanent contracts. It’s not impossible, but as the above poster mentioned, it’s not easy either.
The quality of employees hired was certainly no better than at other companies I’ve worked for where firing was much easier. That’s with a sample size of 1 Dutch company though, and that company sucked, so it’s not a great dataset from which to draw conclusions.
In the financial industry, you don’t get your bonus when you’re on garden leave. This is usually the majority of your pay, so while it’s definitely nice to be paid to do nothing (I’m on garden leave currently until next September), you’re not earning up to your potential.
I’m not sure I understand what you mean. I’ve run HPC clusters for a long time now, and node failures are just a fact of life. If 3 or 4 nodes of your 500 node cluster are down for a few days while you wait for RMA parts to arrive, you haven’t lost much value. Your cluster is still functioning at nearly peak capacity.
You have a handful of nodes that the cluster can’t function without (scheduler, fileservers, etc), but you buy spares and 24x7 contracts for those nodes.
Not in the context the person you responded to meant it. Yes, you can very easily get 50GB/s from a few NVMe devices on a single box. Getting 50GB/s on a POSIX-ish filesystem exported to 1000 servers is very possible and common, but orders of magnitude more complicated. 500GB/s is tougher still. 5TB/s is real tough, but real fun.
This is tangential to your point, but I’ll just mention that Azure has some properly specced out HPC gear: IB, FPGAs, the works. You used to be able to get time on a Cray XC with an Ares interconnect, but I never have occasion to use it, so I don’t know if you still can. They’ve been aggressively hiring top-notch HPC people for a while.
That was definitely one of the weirdest things of working in academia IT: “hey. Can you buy me a workstation that’s as close to $6,328.45 as it is possible to get, and can you do it by 4pm?”
Most of the time, it just that it’s a hassle. It takes a while to prep and tune a big hero run for benchmarking, and if you spend a billion dollars on a cluster, it’s making you a lot more than that. Taking it down for a day or two stops the money printers.