HPC

So, who is the cluster bomb? Student results sliced, diced, analysed

And maybe over-analysed too


HPC Blog It's time to close the books on another highly successful SC Student Cluster Competition. This year was special in a number of ways. First, it was the event's 10th anniversary. At 14 teams, it was also the largest SC competition ever – a far cry from the original five. SC16 was also noteworthy in terms of the performance achieved (more than twice the existing LINPACK record) and the wide variety of cluster configurations designed by the student participants.

We already covered the LINPACK and HPCG results here and here, so let's get right to the applications.

The "Reproducibility" application is a first for any cluster competition. This application has students testing out a scientific paper in order to reproduce the results.

In this test, the students are using ParConnect to analyse pond scum in an exercise of metagenomics. As you can imagine, there are a huge number of different organisms in your typical sample of scum, and assembling their genes is a pretty big compute job.

As you can see from the scores, this application didn't pose a huge challenge to our student teams. The top three teams (USTC, Illinois, and FAU) all scored a perfect 100 per cent on this application, while Utah came in fourth with 98 per cent. Munich and Texas followed with twin 95 per cent scores.

The next application up is Paraview, a visualisation application that is used in a wide variety of fields. It is particularly handy for analysing extremely large datasets, and has been used on supercomputers to analyse petascale-sized datasets.

Team USTC narrowly edged Team Munich on this application by a single point. Running a little bit behind the leaders are Teams Illinois and Utah – which is a very impressive performance for newbie teams. As you can see by the median and mean scores, there was a bit more variability in the scores for the other ten teams.

GROMACS was this year's mystery application – revealed to students just before the application portion of the competition began. Those of you who routinely analyse proteins or lipids, whether for your job or as a hobby, know that GROMACS is one of the mack-daddy programs designed to handle these tasks.

Our students weren't quite as experienced with this program, judging by the average and median scores for this application. Still, the kids from Singapore (Team Nanyang) managed to post a perfect score, followed by the newbie Power-pushing team from Peking at 85 per cent. Team USTC, with their ten P100 accelerators, finished behind Peking by a single point, with USTC finishing in the money with 84 per cent. Illinois gets on the scoreboard with an honourable mention.

The hands-down toughest task for the teams was the Password Cracking challenge. Students are presented with a series of datasets that contain password hashes. It's their task to work through dictionary attacks to decode these hashes into true passwords.

As you can see by the median score, teams were all over the map on this application. This is also the lowest average score of all of the applications in the competition. The organisers really threw the kids a curveball with this one.

Teams were scored by the number of passwords they managed to crack – with Team Taiwan (NTHU) posting the biggest number. Team Utah managed to get within 20 per cent of NTHU's result, which again is a great job by a new team.

But check out Team NEU/Auburn. These are the kids whose cluster was lost in the mail and ended up running a couple of workstations and a single true supercomputer node. Somehow, some way, they were able to grab third place – despite having less and significantly crappier hardware. According to our application experts, it looks like NEU/Auburn was able to find great dictionaries and exploit them to the max in order to achieve this score. Colour me impressed.

The final part of the competition included interviews by HPC experts and the judging of team posters. Each team was interviewed with an eye towards their HPC knowledge, their understanding of their systems, and how well they could explain their optimisations.

Team Munich scored a big win with a 95 per cent score, with Team Texas and Utah knotted at 93 per cent. Teams FAU and Peking were right behind with scores of 91 and 90 respectively.

Overall, the teams did very well on the interview and post part of the competition, with the average and median scores both at 84/85 per cent.

... and the Champion is...

Team USTC topped the other thirteen with an overall score of 89 per cent. They are the first team to ever capture the Overall Championship Award and the Highest LINPACK Award at the same competition – an amazing performance.

What's interesting is that they only won two of the application categories (along with LINPACK and HPCG). This was a very close competition. USTC finished in the money on GROMACS, but didn't make the scoreboard on the other applications. I think the big key to their win was not making any mistakes. They turned in an all-round good performance, while posting wins on Paraview and Reproducibility.

Look at Team Utah! They're both a first-time competitor and the home team – and they brought home the Silver Medal (or, since they were already at home, they just sort of kept it there). They built their second place score by placing on the scoreboard on four out of the five applications, plus doing well on the benchmarks.

This is unprecedented in SC competitions. Typically the home team doesn't do so well, and particularly if they're first-timers. This was a fantastic performance by the Utes.

Veterans NTHU took down third place, fuelled by their dominant performance on Password Cracking and solid scores on the other applications/benchmarks. This is the highest the team has placed for several competitions. Perhaps we're seeing the start of a resurgence in Taiwan student clustering?

The blended team of University of Texas (Austin) and Texas State University grabbed an Honourable Mention for finishing only two points south of third place, the best performance we've seen with a blended team in competition history. Quite an achievement.

This competition was unusual in that all of the teams did pretty well. There weren't a lot of zero scores on the scorecards, which is a break from what we've seen in the past. Students are getting better, the hardware is getting better, and the scores are rising. Things are going well in Student Cluster land.

The next major international competition is scheduled for Asia sometime around April. We'll be there to cover all the action and bring you the most comprehensive overview and results possible. Stay tuned.

Similar topics


Other stories you might like

  • Florida's content-moderation law kept on ice, likely unconstitutional, court says
    So cool you're into free speech because that includes taking down misinformation

    While the US Supreme Court considers an emergency petition to reinstate a preliminary injunction against Texas' social media law HB 20, the US Eleventh Circuit Court of Appeals on Monday partially upheld a similar injunction against Florida's social media law, SB 7072.

    Both Florida and Texas last year passed laws that impose content moderation restrictions, editorial disclosure obligations, and user-data access requirements on large online social networks. The Republican governors of both states justified the laws by claiming that social media sites have been trying to censor conservative voices, an allegation that has not been supported by evidence.

    Multiple studies addressing this issue say right-wing folk aren't being censored. They have found that social media sites try to take down or block misinformation, which researchers say is more common from right-leaning sources.

    Continue reading
  • US-APAC trade deal leaves out Taiwan, military defense not ruled out
    All fun and games until the chip factories are in the crosshairs

    US President Joe Biden has heralded an Indo-Pacific trade deal signed by several nations that do not include Taiwan. At the same time, Biden warned China that America would defend Taiwan from attack; it is home to a critical slice of the global chip industry, after all. 

    The agreement, known as the Indo-Pacific Economic Framework (IPEF), is still in its infancy, with today's announcement enabling the United States and the other 12 participating countries to begin negotiating "rules of the road that ensure [US businesses] can compete in the Indo-Pacific," the White House said. 

    Along with America, other IPEF signatories are Australia, Brunei, India, Indonesia, Japan, South Korea, Malaysia, New Zealand, the Philippines, Singapore, Thailand and Vietnam. Combined, the White House said, the 13 countries participating in the IPEF make up 40 percent of the global economy. 

    Continue reading
  • 381,000-plus Kubernetes API servers 'exposed to internet'
    Firewall isn't a made-up word from the Hackers movie, people

    A large number of servers running the Kubernetes API have been left exposed to the internet, which is not great: they're potentially vulnerable to abuse.

    Nonprofit security organization The Shadowserver Foundation recently scanned 454,729 systems hosting the popular open-source platform for managing and orchestrating containers, finding that more than 381,645 – or about 84 percent – are accessible via the internet to varying degrees thus providing a cracked door into a corporate network.

    "While this does not mean that these instances are fully open or vulnerable to an attack, it is likely that this level of access was not intended and these instances are an unnecessarily exposed attack surface," Shadowserver's team stressed in a write-up. "They also allow for information leakage on version and build."

    Continue reading
  • A peek into Gigabyte's GPU Arm for AI, HPC shops
    High-performance platform choices are going beyond the ubiquitous x86 standard

    Arm-based servers continue to gain momentum with Gigabyte Technology introducing a system based on Ampere's Altra processors paired with Nvidia A100 GPUs, aimed at demanding workloads such as AI training and high-performance compute (HPC) applications.

    The G492-PD0 runs either an Ampere Altra or Altra Max processor, the latter delivering 128 64-bit cores that are compatible with the Armv8.2 architecture.

    It supports 16 DDR4 DIMM slots, which would be enough space for up to 4TB of memory if all slots were filled with 256GB memory modules. The chassis also has space for no fewer than eight Nvidia A100 GPUs, which would make for a costly but very powerful system for those workloads that benefit from GPU acceleration.

    Continue reading
  • GitLab version 15 goes big on visibility and observability
    GitOps fans can take a spin on the free tier for pull-based deployment

    One-stop DevOps shop GitLab has announced version 15 of its platform, hot on the heels of pull-based GitOps turning up on the platform's free tier.

    Version 15.0 marks the arrival of GitLab's next major iteration and attention this time around has turned to visibility and observability – hardly surprising considering the acquisition of OpsTrace as 2021 drew to a close, as well as workflow automation, security and compliance.

    GitLab puts out monthly releases –  hitting 15.1 on June 22 –  and we spoke to the company's senior director of Product, Kenny Johnston, at the recent Kubecon EU event, about what will be added to version 15 as time goes by. During a chat with the company's senior director of Product, Kenny Johnston, at the recent Kubecon EU event, The Register was told that this was more where dollars were being invested into the product.

    Continue reading
  • To multicloud, or not: Former PayPal head of engineering weighs in
    Not everyone needs it, but those who do need to consider 3 things, says Asim Razzaq

    The push is on to get every enterprise thinking they're missing out on the next big thing if they don't adopt a multicloud strategy.

    That shove in the multicloud direction appears to be working. More than 75 percent of businesses are now using multiple cloud providers, according to Gartner. That includes some big companies, like Boeing, which recently chose to spread its bets across AWS, Google Cloud and Azure as it continues to eliminate old legacy systems. 

    There are plenty of reasons to choose to go with multiple cloud providers, but Asim Razzaq, CEO and founder at cloud cost management company Yotascale, told The Register that choosing whether or not to invest in a multicloud architecture all comes down to three things: How many different compute needs a business has, budget, and the need for redundancy. 

    Continue reading

Biting the hand that feeds IT © 1998–2022