ExaScale is a 4-way Competition

In this post-ISC show, the RadioFree team discusses

• Magical cooling technology from Europe. Dan goes over magic beads that draw heat away and can carry-on doing it pretty much forever in a technology from the venerable Fraunhofer Institute and showcased by Lenovo.
• How pursuit of ExaScale computing is turning into heated competition with the US, China, Japan, and Europe. The European effort is targeting 2 pre-exa installation in the coming months, and 2 actual ExaScale installations in the 2022-2023 timeframe at least one of which will be based on European technology. This presumably refers to the European Process Initiative.
  The software ecosystem is an important consideration and how they all evolve and whether or not they converge will be a big issue.
• Another heated competition at the ISC Student Cluster Competition with the team from South Africa claiming the top spot. Dan has developed an efficiency metric that he will unveil in a future episode. This could separate the prowess of the team from that of the system!

Catch of the Week

Henry:

Henry point out the challenge for customers when the company that breached their data goes out of business.

Collections Firm Behind LabCorp, Quest Breaches Files for Bankruptcy

A medical billing firm responsible for a recent eight-month data breach that exposed the personal information on nearly 20 million Americans has filed for bankruptcy, citing “enormous expenses” from notifying affected consumers and the loss of its four largest customers.

Shahin:

Shahin highlights a paper on the beginnings of the programming language APL. A cool historical account.

The Socio-Technical Beginnings of APL, by Eugene McDonnell

This paper gives some of the history of implementations of APL, and concentrates on the system aspects of these implementations, paying special attention to the evolution of the workspace concept, the time-sharing scheduling strategy, and the handling of the terminal. It contrasts the development of APL with the development of other time-sharing systems which were being built at the same time.

Dan:

Dan relays the sad story of the multi-year demise of a the honor bar at the hotel near ISC.
Listen in to hear the full conversation.

Download the MP3 * Subscribe on iTunes * RSS Feed

Sign up for our insideHPC Newsletter

Why did HPE buy Cray?

Why did HPE buy Cray?

The RFHPC team tackles the HPE-Cray acquisition as it reviews the companies' recent moves and strengths and market conditions in the context of:

• the 5-tier data center application architecture: Embedded, Mobile, Desktop, On-premises, Off-premises
• the emergence of AI as a must-do enterprise app, and
• increasing commonality between supercomputers and enterprise servers.

Catch of the Week

Henry:

Another week another breach!

Massive Quest Diagnostics data breach impacts 12 million patients

A massive data breach has struck Quest Diagnostics and the information of up to 11.9 million patients has potentially been compromised. On Monday, the US clinical laboratory said that American Medical Collection Agency (AMCA), a billing collections provider that works with Quest, informed the company that an unauthorized user had managed to obtain access to AMCA systems.

Dan:

Dan points out that the new Apple Mac Pro can be configured to cost tens of thousands of dollars. Given that he and Dan are PC people, the nuances of the Apple value are obviously lost of them, goes the counter argument.

Apple’s top spec Mac Pro will likely cost at least $35,000

That’s before you count the GPUs or a Pro Display XDR screen.
Apple announced today that its new Mac Pro starts at an already pricey $6,000, but the company neglected to mention how much the top-of-the-line model will cost. So we shopped around for equivalent parts to the top-end spec that Apple’s promising. As it turns out: $33,720.88 is likely the bare minimum — and that’s before factoring in the four GPUs, which could easily jack that price up to around $45,000.

Listen in to hear the full conversation.

Download the MP3 * Subscribe on iTunes * RSS Feed

Sign up for our insideHPC Newsletter

Quantum Computing and HPC

Quantum Computing and HPC

Another scintillating and insightful episode of RFHPC is about Quantum Computing and HPC and how the two spaces are evolving and cooperating.

We welcome a a distinguished guest with a most suitable background to talk to us about HPC and Quantum Computing. Mike Booth,  who’s been in supercomputing since 1979 including stints at Cray through 2000 where he ran the Software and Applications division and was later a GM at StorageTek heading the network storage division. He got into Quantum Computing when he joined D-Wave. He had just accepted to be the CTO of Quantum Computing, Inc. when we recorded this show.

We discuss and touch on how Quantum Computing and HPC interface, analog vs digital, qubits, magnets, resistors, connectors, cryogenics, algorithms, languages, the huge search spaces, NP-complete problems, quadratic unconstrained binary optimization (Qubo), Tabu search, etc. and how they are two different games right now but touching two sides of the big problems that represent grand challenges. Because QC is an accelerator, it fits nicely with how a lot of HPC is being done today.

We’re going to have to bring Mike back and we look forward to that.

ExaScale at Oakridge

Mike happens to be in Tennessee, and the episode was recorded when the new ExaScale system at Oakridge was announced so the team. That was quite a significant day for US science, and a second big win for Cray, this time with AMD. It's one of the few large systems that is not based on Intel or Nvidia technologies, and was described as:

• 100 Cray Shasta cabinets
• 40 MW power
• More than 1 million lbs weight
• 7,300 square feet
• 90 miles of cabling
• 5,900 gallons of water per minute for cooling

We don't remember who exactly had a hard stop, but no time for Catch of the Week this week, which some of you would be pleased to hear!

Give it a listen (and take good notes!)

Download the MP3 * Subscribe on iTunes * RSS Feed

Sign up for our insideHPC Newsletter

TOP500 Jun2019, Facebook Coin

The new TOP500 list of most powerful supercomputers is out and we do our usual quick analysis. Not much changed in the TOP10 but a lot is changing further down the list. Here is a quick take:

• There are 65 new entries in 2019.
• US science is receiving support via DOE sites and academic sites like TACC.
• 26 countries are represented. China continues to widen its lead, now with 219 entries, followed by the US with 116, Japan with 29, France with 19, the UK with 18, Germany with 14, Ireland and the Netherlands with 13 each, and Singapore with 10.
• Vendors substantially reflect the country standings. Lenovo has 175 entries, Inspur 71, and Sugon 63, all in China. Cray with 42 and HPE with 40 (which will combine when their deal closes), followed by Dell at 17 and IBM at 16.  Bull has 21 entries.
• There are a lot of "accidental supercomputers" on the list. These are systems that probably are not be doing much science or AI work but they could, and the vendors counted them and it seems to be within the rules to list them. It's controversial but not a new practice.
• There are several systems listed as "Internet" companies. Hard to tell what that means but it points to the existence of very large clusters in the cloud for whatever purpose. Last year, there was one system listed as Amazon EC2, which remains on the list. This time, there is also one at Facebook. Usually the big social/cloud players don't care to participate, though they obviously could summon the resources to run the benchmarks.
• Just over half of systems use Ethernet as a fabric. A quarter us InfiniBand, nearly 50 use Intel's OmniPath, and the rest, 55, use custom interconnects like the ones Cray provides. The team talks about Cray+HPE entering the interconnect business for real and if so, they will be formidable.
• The majority of entries, 367, do not have any accelerators. 125 use Nvidia GPUs.
• The overwhelming majority of the systems, 478 of them, are based on Intel CPUs. 13 are IBM, and there is 1 system based on Arm provided by Cavium, now part of Marvell.
• So the when it comes to chips, it's an Intel game with a respectable showing by Nvidia when GPUs are used. Alternatives are bound to appear as the tens and tens of AI chips in the works become available and Arm, AMD, and IBM build on. The recently announced system at Oakridge will be all AMD, and that will point to an alternative as well.
• Notably, Intel is listed as the vendor for 2 entries and Nvidia is listed for 4. While Intel has stayed largely away from looking like a system vendor, Nvidia is going for it with its usual alacrity. That, and the pending acquisition of Mellanox by Nvidia should serve as a warning to all system vendors who might feel stuck between treating Nvidia as an important supplier and an up and coming competitor.

CryptoSuper500

Shahin mentions the 2nd edition of the CryptoSuper500 list (really 50 for now), a list developed by his colleague Dr. Stephen Perrenod, which was launched last November, and is being released at the same time as the TOP500. The TOP500 has spawned variations that look at different workloads and attributes, for example, the Green500, Graph500, and IO500 lists. CryptoSuper500 was inspired by those lists. The material for the inaugural edition of the CryptoSuper500 list here.
Cryptocurrency mining operations are often pooled and are very much supercomputing class, typically using accelerator technologies such as custom ASICs, FPGAs, or GPUs. Bitcoin is the most notable of such currencies. Scroll down for the top-10 list and see the slides for the full list and the methodology.

Catch of the Week

Henry:

Henry talks about check-out lanes at Target all being down for unknown reasons, though he hesitates to call that a cybersecurity breach. It turned out he's right and the company blamed an "internal technology issue".

Target down (then back up) as cash registers fail and leave long lines

Target's payment systems appeared to be missing the mark the day before Father's Day, as terminals went AWOL for a couple of hours in a number of the company's US retail outlets. The outage caused long lines but prompted an encouraging show of sympathy for Target employees from people on Twitter. And there were some jokes too, of course.

Shahin:

Facebook is expected to release a new cryptocurrency that is already impacting the crypto market.

Here’s what we know so far about the secretive Facebook coin

Facebook is likely to release information about its secretive cryptocurrency project, codenamed Libra, as soon as June 18, TechCrunch reports.

As is traditional with new cryptocurrencies, the social networking giant is expected to release a so-called “white paper” outlining how the currency works and the company’s plans for it.

Dan:

Dan reminds us all of the inimitable Erich Anton Paul von Däniken and his ancient astronauts hypotheses!

Listen in to hear the full conversation.

Download the MP3 * Subscribe on iTunes * RSS Feed

Sign up for our insideHPC Newsletter

Forty+ different AI chips

What are we going to do with 40+ different AI chips?

This week, the team looks at AI chips again, this time motivated by an article in EE Times about once such chip, Graphcore, and touts it as "the most complex processor" ever at some 20 billion transistors. The VC-backed company out of Bristol, UK is also valued on paper at $1.7b, gaining it the coveted "unicorn" status, apparently the "only western semi-conductor unicorn".

This being one of 40+ such AI chips (and that may be conservative), the odds of success are tough and the task formidable. But even if only 2 or 3 of such chips become successful, that's already a significant disruption to the market.

The Graphcore chip is 16nm, 1.6GHz, and comes in a PCIe card at 300W. You can stack 8 of these in a 4U chassis, so 2.4 kW just for those.

After a mini-rant about respected publications succumbing to clickbaits, the team talks about how cooling will be an issue and calls again for more clarity in performance metrics since the chip is rated at 125 TFlops but we don't know at what precision. Shahin reminds the team of his suggestion to clarify things by including precision in the metric, like DFlops for double precision, and then S for single, H for half, and Q for quarter precision.

Henry talks about how hard it is to build and test complex software like this despite Shahin's view that the modern software stack is too high so the chip need only be concerned with a couple of layers, codes are new and open to getting recompiled, it's increasingly open source, cloud providers and large customers have the wherewithal to do the job, and traditional HPC customers have the willingness to do the work if performance enhancements are there.

No "Catch of the Week" this time since Henry had a hard stop. We're used to it!

Listen in to hear the full conversation.

Download the MP3 * Subscribe on iTunes * RSS Feed

Sign up for our insideHPC Newsletter

Amdahl's Law and GPUs, Asian Student Cluster Competition

Results of the Asian Student Cluster Competition

In this episode, Dan has just come back from China and reviews the results of the Asian Student Cluster Competition and HPC workshop.

For the first time, a non-mainland-Chinese team wins the top spot. Taiwan takes the gold in part by their stellar performance in HPCG benchmark where they achieved 2 TFlops, some 25% better than the 2nd best team. The system was a 5-node cluster with Infiniband FDR interconnect. Other interesting info is shared on various codes and configurations.

GPUs and Amdahl's Law

Dan also mentions that reports from some of the TOP500 sites suggest that GPUs are doing 93-97% of the computation. This sounds very impressive but Shahin points out that since GPUs have hundreds of cores, they should be doing much better, that 93-97% is in fact not as good as it should be at that scale of system and problem size. He is still waiting for some actual utilization data on GPUs too.

Catch of the Week

Henry:

Henry points out many security cameras, offered by several brands but are all manufactured by the same vendor back in China, have big time vulnerabilities so he's staying away from all of them until further notice. Shahin wonders why they are called "security" cameras!

P2P Weakness Exposes Millions of IoT Devices

A peer-to-peer (P2P) communications technology built into millions of security cameras and other consumer electronics includes several critical security flaws that expose the devices to eavesdropping, credential theft and remote compromise, new research has found.

Shahin:

Shahin talks about Jaguar-Land Rover planning to offer a cryptocurrency wallet to reward drivers that participate in providing traffic and other types of data. He likes their catch phrase: zero emission, zero accident, zero congestion.

ON THE MONEY: EARN AS YOU DRIVE WITH JAGUAR LAND ROVER

Drivers will be able to earn cryptocurrency and make payments on the move using innovative connected car services being tested by Jaguar Land Rover.

Dan:

Dan laments the confiscation of his external camera battery at the airport in China because the spec label was a little worn off and the authorities could not read it to ascertain its safety despite his willingness to get a note from the airline, etc.  Nice expensive battery, but at a medium-sized paperback book, maybe following the rules strictly is not a bad idea.

Listen in to hear the full conversation.

Download the MP3 * Subscribe on iTunes * RSS Feed

Sign up for our insideHPC Newsletter

In-Memory, Caching, Persistence

Asian Student Cluster Competition

The ever-popular "calendar segment" is back (yes, that is us being sarcastic) as Dan describes his upcoming trip to Dalian, China for the Asian Student Cluster competition. 20 teams are expected to compete, mostly from mainland China, but also from Europe, South America and other parts of Asia. All of them have gone through a rigorous qualifying process. Chinese vendor Inspur provides all the equipment (except for accelerators, which remain the responsibility of the teams) based on the configuration that the teams request. As always, the team, and the rest of humanity, is highly supportive of the student cluster tournaments that take place at SC, ISC, and this Asian one, in addition to many regional supporting events. Dan also lets us know how to go about picking the right seat in the airplane. There's an app for that!

New Optane

Intel’s new Optane is M.2 format with 16GB or 32GB of Opatne and 256GB, 512GB, or 1TB of Flash backing it. This made Henry think of history of caching and how people used to go through laborious data placement on outer cylinders of hard disks. Then you had DIMMs with Flash behind it. The discussion goes to the algorithms and policies that manage data movement, which is another form of optimizing workload management, and the different combinations of fast/small/expensive capacity backed up with slower/larger/cheaper capacity. And how persistence impacts this equation and to what extent applications may want to optimize around these. Then there is in-memory processing, ram-disk, and how in the future, the ever-increasing size of memory can make those standard practice.

Catch of the Week

Henry:

Henry is catchless but we can expect a healthy debate. Shahin produces two to make up for it.

Shahin:

Shahin has new reinforcement that IoT and HPC will converge in interesting ways. 1) We all know that IoT is the fountain of data and will generates so much data that you need HPC (and AI) to make sense of them, but 2) what is also very interesting Digital twins will need to simulate and predict the behavior of the real thing.

What we mean when we talk about digital twins

"Watching it prompted me to wonder how much data it takes to create a digital twin. If I had a digital simulacrum of a machine and could apply different environmental or mechanical factors to it, how large could that original simulacrum be? It turns out that’s not how digital twins work. They aren’t virtual doppelgangers. They are actually a series of algorithms that connote how machine moves or behaves. In other words, a twin isn’t a twin so much as it’s…math."

Amazon Workers Are Listening to What You Tell Alexa

"A global team reviews audio clips in an effort to help the voice-activated assistant respond to commands. The team comprises a mix of contractors and full-time Amazon employees who work in outposts from Boston to Costa Rica, India and Romania, according to the people, who signed nondisclosure agreements barring them from speaking publicly about the program. They work nine hours a day, with each reviewer parsing as many as 1,000 audio clips per shift, according to two workers based at Amazon’s Bucharest office, which takes up the top three floors of the Globalworth building in the Romanian capital’s up-and-coming Pipera district. The modern facility stands out amid the crumbling infrastructure and bears no exterior sign advertising Amazon’s presence."

Shahin thinks it’s as simple as supervised learning needing supervisors, so the question is the legal framework and jurisdiction issues, and social policy, and not technology.

Dan:

It is time to panic more as Dan shares the story of the Chinese scientist who’s presumably blended monkey and human genes. It’s Planet of the Apes all over again.

Chines Scientists Gene-Hacked Super Smart Human-Monkey Hybrids

"For the first time, scientists have used gene-editing techniques to make monkey brains more humanlike. The monkeys, rhesus macaques, got smarter — they had superior memories to unaltered monkeys, according to recently-published research that’s kicked off a fiery debate among ethicists about how far scientists should be able to take genetic experimentation."

Listen in to hear the full conversation.
Download the MP3 * Subscribe on iTunes * RSS Feed Sign up for our insideHPC Newsletter