August 22, 2017 Ken Strandberg
Professor Satoshi Matsuoka, of the Tokyo Institute of Technology (Tokyo Tech) researches and designs large scale supercomputers and similar infrastructures. More recently, he has worked on the convergence of Big Data, machine/deep learning, and AI with traditional HPC, as well as investigating the Post-Moore Technologies towards 2025.
He has designed supercomputers for years and has collaborated on projects involving basic elements for the current and more importantly future Exascale systems. I talked with him recently about his work with the Tsubame supercomputers at Tokyo Tech. This is the first in a two-part article. For background on the Tsubame 3 system we have an in-depth article here from earlier this year.
TNP: Your new Tsubame 3 supercomputer is quite a heterogeneous architecture of technologies. Have you always built heterogeneous machines?
Satoshi Matsuoka, professor of the High Performance Computing Systems Group, GSIC, at the Tokyo Institute of Technology, showing off the Tsubame 3.0 server node.
Matsuoka: Ive been building clusters for about 20 years nowfrom research to production clustersall in various generations, sizes, and forms. We built our first very large-scale production cluster for Tokyo Techs supercomputing center back in 2006. We called it Tsubame 1, and it beat the then fastest supercomputer in Japan, the Earth Simulator.
We built Tsubame 1 as a general-purpose cluster, instead of a dedicated, specialized system, as the Earth Simulator was. But, even as a cluster it beat the performance in various metrics of the Earth Simulator, including the top 500, for the first time in Japan. It instantly became the fastest supercomputer in the country, and held that position for the next 2 years.
I think we are the pioneer of heterogeneous computing. Tsubame 1 was a heterogenous cluster, because it had some of the earliest incarnations of accelerators. Not GPUs, but a more dedicated accelerator called Clearspeed. And, although they had a minor impact, they did help boost some application performance. From that experience, we realized that heterogeneous computing with acceleration was the way to go.
TNP: You seem to also be a pioneer in power efficiency with three wins on the Green 500 list. Congratulations. Can you elaborate a little about it?
Matsuoka: As we were designing Tsubame 1, it was very clear that, to hit the next target of performance for Tsubame 2, which we anticipated would come in 2010, we would also need to plan on reducing overall power. Weve been doing a lot of research in power-efficient computing. At that time, we had tested various methodologies for saving power while also hitting our performance targets. By 2008, we had tried using small, low-power processors in lab experiments. But, it was very clear that those types of methodologies would not work. To build a high-performance supercomputer that was very green, we needed some sort of a large accelerator chip to accompany the main processor, which is x86.
We knew that the accelerator would have to be a many core architecture chip, and GPUs were finally becoming usable as a programming device. So, in 2008, we worked with Nvidia to populate Tsubame 1 with 648 third-generation Tesla GPUs. And, we got very good results on many of our applications. So, in 2010, we built Tsubame 2 as a fully heterogenous supercomputer. This was the first peta-scale system in Japan. It became #1 in Japan and #4 in the world, proving the success of a heterogeneous architecture. But, it was also one of the greenest machines at #3 on the Green 500, and the top production machine on the Green 500. The leading two in 2010 were prototype machines. We won the Gordon Bell prize in 2011 for the configuration, and we received many other awards and accolades.
It was natural that when we were designing Tsubame 3, we would continue our heterogeneous computing and power efficiency efforts. So, Tsubame 3 is the second-generation, large-scale production heterogeneous machine at Tokyo Tech. It contains 540 nodes, each with four Nvidia Tesla P100 GPUs (2,160 total), two 14-core Intel Xeon Processor E5-2680 v4 (15,120 cores total), two dual-port Intel Omni-Path Architecture (Intel OPA) 100 Series host fabric adapters (2,160 ports total), and 2 TB of Intel SSD DC Product Family for NVMe storage devices, all in an HPE Apollo 8600 blade, which is smaller than a 1U server.
A lot of the enhancements that went into the machine are specifically to make it a more efficient machine as well as for high performance. The result is that Tsubame 3although at the time of measurement for the June 2017 lists we only ran on a small subset of the full configurationis #61 on the Top 500 and #1 on the Green 500 with 14.11 gigaflops/watt, an RMax of just under 2 petaflops, and a theoretical peak of over 3 petaflops. Tsubme 3 just became operational August 1, with its full 12.1 petaflops configuration, and we hope to have the scores for the full configuration for the November benchmark lists, including the Top500 and the Green500.
TNP: Tsubame 3 is not only a heterogeneous machine, but built with a novel interconnect architecture. Why did you choose the architecture in Tsubame 3?
Matsuoka : That is one area where Tsubame 3 is different, because it touches on the design principles of the machine. With Tsubame 2, many applications experienced bottlenecks, because they couldnt fully utilize all the interconnect capability in the node. As we were designing Tsubame 3, we took a different approach. Obviously, we were planning on a 100-gigabit inter-node interconnect, but we also needed to think beyond just speed considerations and beyond just the node-to-node interconnect. We needed massive interconnect capability, considering we had six very high-performance processors that supported a wide range of workloads, from traditional HPC simulation to big data analytics and artificial intelligence, all potentially running as co-located workloads.
For the network, we learned from the Earth simulator back in 2002 that to maintain application efficiency, we needed to sustain a good ratio between memory bandwidth and injection bandwidth. For the Earth Simulator, that ratio was about 20:1. So, over the years, Ive tried to maintain a similar ratio in the clusters weve built, or set 20:1 as a goal if it was not possible to reach it. Of course, we also needed to have high bisection bandwidth for many workloads.
Todays processors, both the CPUs and GPUs, have significantly accelerated FLOPS and memory bandwidth. For Tsubame 3, we were anticipating certain metrics of memory bandwidth in our new GPU, plus, the four GPUs were connected in their own network. So, we required a network that would have a significant injection bandwidth. Our solution was to use multiple interconnect rails. We wanted at least one 100 gigabit injection port per GPU, if not more.
For high PCIe throughput, instead of running everything through the processors, we decided to go with a direct-attached architecture using PCIe switches between the GPUs, CPUs, and Intel OPA host adapters. So, we have full PCIe bandwidth between all devices in the node. Then, the GPUs have their own interconnect between themselves. Thats three different interconnects within a single node.
If you look at the bandwidth of these links, theyre not all that different. Intel OPA is 100 gigabits/s, or 12.5 GB/s. PCIe is 16 GB/sec. NVLink is 20 GB/sec. So, theres less than a 2:1 difference between the bandwidth of these links. As much as possible we are fully switched within the node, so we have full bandwidth point to point across interconnected components. That means that under normal circumstances, any two components within the system, be it processor, GPU, or storage, are fully connected at a minimum of 12.5 GB/sec. We believe that this will serve our Tsubame 2 workloads very well and support new, emerging applications in artificial intelligence and other big data analytics.
TNP: Why did you go with the Intel Omni-Path fabric?
Matsuoka: As I mentioned, we always focus on power as well as performance. With a very extensive fabric and a high number of ports and optical cables, power was a key consideration. We worked with our vendor, HPE, to run many tests. The Intel OPA host fabric adapter proved to run at lower power compared to InfiniBand. But, as important, if not more important, was thermal stability. In Tsubame 2, we experienced some issues around interconnect instability over its long operational period. Tsubame 3 nodes are very dense with a lot of high-power devices, so we wanted to be make sure we had a very stable system.
A third consideration was Intel OPAs adaptive routing capability. Weve run some of our own limited-scale tests. And, although we havent tested it extensively at scale, we saw results from the University of Tokyos very large Oakforest-PACS machine with Intel OPA. Those indicate that the adaptive routing of OPA works very, very well. And, this is critically important, because one of our biggest pain points of Tsubame 2 was the lack of proper adaptive routing, especially when dealing with degenerative effects of optical cable aging. Over time AOCs die, and there is some delay between detecting a bad cable and replacing it or deprecating it. We anticipated Intel OPA, with its end-to-end adaptive routing, would help us a lot. So, all of these effects combined gave the edge to Intel OPA. It wasnt just the speed. There were many more salient points by which we chose the Intel fabric.
TNP: With this multi-interconnect architecture, will you have to do a lot of software optimization for the different interconnects?
Matsuoka: In an ideal world, we will have only one interconnect, everything will be switched, and all the protocols will be hidden underneath an existing software stack. But, this machine is very new, and the fact that we have three different interconnects is reflecting the reality within the system. Currently, except for very few cases, there is no comprehensive catchall software stack to allow all of these to be exploited at the same time. There are some limited cases where this is covered, but not for everything. So, we do need the software to exploit all the capability of the network,including turning on and configuring someappropriate DMA engines, or some pass through, because with Intel OPA you need some CPUs involvement for portions of the processing.
So, getting everything to work in sync to allow for this all-to-all connectivity will require some work. Thats the nature of the research portion of our work on Tsubame 3. But, we are also collaborating with people like a team at The Ohio State University.
We have to work with some algorithms to deal with this connectivity, because it goes both horizontally and vertically. The algorithms have to adapt. We do have several ongoing works, but we need to generalize this to be able to exploit the characteristics of both horizontal and vertical communications between the nodes and the memory hierarchy. So far, its very promising. Even out of the box, we think the machine will work very well. But as we enhance our software portion of the capabilities, we believe the efficiency of the machine will become higher as we go along.
In the next article in this two-part series later this week, Professor Matsuoka talks about co-located workloads on Tsubame 3.
Ken Strandberg is a technical story teller. He writes articles, white papers, seminars, web-based training, video and animation scripts, and technical marketing and interactive collateral for emerging technology companies, Fortune 100 enterprises, and multi-national corporations. Mr. Strandbergs technology areas include Software, HPC, Industrial Technologies, Design Automation, Networking, Medical Technologies, Semiconductor, and Telecom. He can be reached at ken@catlowcommunications.com.
Categories: HPC
Tags: OmniPath, Tsubame
An Early Look at Baidus Custom AI and Analytics Processor First In-Depth View of Wave Computings DPU Architecture, Systems
View post:
Inside View: Tokyo Tech's Massive Tsubame 3 Supercomputer - The Next Platform
- New Microsoft Ads Take Aim at Mac Pricing - November 8th, 2009 [November 8th, 2009]
- Adobe Flash Comes to TV - November 8th, 2009 [November 8th, 2009]
- Microsoft Introduces Windows 7 Starter Edition - November 8th, 2009 [November 8th, 2009]
- Mac Viruses and Trojans Becoming More Prevalent - November 8th, 2009 [November 8th, 2009]
- Apple ‘Customer Experience’ Continues to Trounce PCs - November 8th, 2009 [November 8th, 2009]
- Seagate Introduces ‘Replica’ Drive to Backup Entire PC - November 8th, 2009 [November 8th, 2009]
- Still Love XP? Run it on Windows 7! - November 8th, 2009 [November 8th, 2009]
- Is Microsoft Ditching Vista? - November 8th, 2009 [November 8th, 2009]
- The Kindle DX: Not Exactly a Textbook Killer - November 8th, 2009 [November 8th, 2009]
- The Smart Shopper’s Guide to Buying a Wireless Router - May 19th, 2010 [May 19th, 2010]
- iTunes 10: So Long, Ringtone Creator - Thanks for the Memories - October 17th, 2010 [October 17th, 2010]
- iTunes 10: So Long, Ringtone Creator – Thanks for the Memories - February 14th, 2011 [February 14th, 2011]
- How to Make Your Laptop Last Longer - February 14th, 2011 [February 14th, 2011]
- Client Build 5 UPDATE: Personal Super Computer 2011 (SR-2 X5690 OCZ Vertex 3 GTX590 Nvidia Tesla) - Video - March 29th, 2012 [March 29th, 2012]
- Super Micro Computer, Inc. Announces 3rd Quarter 2012 Financial Results - April 25th, 2012 [April 25th, 2012]
- Super Micro Computer Q3 Profit Slips - Quick Facts - April 25th, 2012 [April 25th, 2012]
- Super Computer Maker Cray and Intel strike Partnership - April 25th, 2012 [April 25th, 2012]
- Super Micro Computer Q3 12 Earnings Conference Call At 5:00 PM ET - April 25th, 2012 [April 25th, 2012]
- Herd mentallity and the information super highway - Video - April 25th, 2012 [April 25th, 2012]
- Brain vs. Computer - Video - May 4th, 2012 [May 4th, 2012]
- Minecraft World First - Most wanted redstone device - Video - May 4th, 2012 [May 4th, 2012]
- PS3 Jailbreak Tutorial 4.11 WORKING - Video - May 4th, 2012 [May 4th, 2012]
- China's Tianhe-1 supercomputer begins operations - Video - May 4th, 2012 [May 4th, 2012]
- June 2011 TOP500 Review looks at Japan's K Supercomputer - Video - May 4th, 2012 [May 4th, 2012]
- Super Vision for Soldiers - May 5th, 2012 [May 5th, 2012]
- The Super Sonic Show Episode 0-Computer Help - Video - May 7th, 2012 [May 7th, 2012]
- Why Super Micro Computer's Earnings May Be Less Than Awesome - May 10th, 2012 [May 10th, 2012]
- Magnetic bacteria may help build computer hard drives - May 10th, 2012 [May 10th, 2012]
- SUPER WHY! Around the World Adventure Kicks off PBS KIDS Summer Learning Initiative This June - May 10th, 2012 [May 10th, 2012]
- Tutorial SUPER COMPUTER girl 3750 sylvia Vs fem game 4 (3550) - Video - May 10th, 2012 [May 10th, 2012]
- SUPER COMPUTER Wii best 3750 sylvia Vs learn chess 4 (3550) - Video - May 10th, 2012 [May 10th, 2012]
- SUPER COMPUTER girls city 3750 sylvia Vs RYBKA 4 (3550) - Video - May 10th, 2012 [May 10th, 2012]
- John Laban - Open University Super Computer Room - Video - May 10th, 2012 [May 10th, 2012]
- Can A Super Computer Save Banking? Part 2 of 2 - Video - May 10th, 2012 [May 10th, 2012]
- Supermicro® Launches Widest Range of UP Server Platforms Supporting Intel® Xeon® E3-1200 v2 - May 16th, 2012 [May 16th, 2012]
- Supermicro® Debuts New X9 DP and 4-Way MP Platforms - May 16th, 2012 [May 16th, 2012]
- Supermicro® Launches Widest Range of Server Platforms Supporting Intel® Xeon® E3-1200 v2 - May 16th, 2012 [May 16th, 2012]
- Invention kit for banana pianos, alphabet soup keyboards - May 16th, 2012 [May 16th, 2012]
- A few errors could be key to super-efficient computer chips - May 20th, 2012 [May 20th, 2012]
- Supermicro® Highlights Latest GPU SuperServer®, SuperBlade® and ... - May 20th, 2012 [May 20th, 2012]
- Kontron HPEC Platform Chosen by Military Embedded Systems Magazine for Editor's Choice Award - May 20th, 2012 [May 20th, 2012]
- Raspberry Pi to rebirth an era of Woz-like super creativity? - May 20th, 2012 [May 20th, 2012]
- Taste and tale of success - May 20th, 2012 [May 20th, 2012]
- 1 Reason to Expect Big Things From Super Micro Computer - May 25th, 2012 [May 25th, 2012]
- Bump's Super Popular App Just Got A Million Times Cooler With Its Latest Update - May 25th, 2012 [May 25th, 2012]
- Is The Computer 'Cloud' Compromising You Privacy? - May 26th, 2012 [May 26th, 2012]
- Super MP3 Download 4.8.2.6 - May 28th, 2012 [May 28th, 2012]
- Radiohead's Kid A and OK Computer, Now in 8-Bit - May 29th, 2012 [May 29th, 2012]
- ASUS P6T7 WS Super Computer MoBo - Video - May 29th, 2012 [May 29th, 2012]
- Photonic Super Computer 2012 - Video - May 29th, 2012 [May 29th, 2012]
- Kaspersky discovers super-complex Flame malware - May 30th, 2012 [May 30th, 2012]
- Supermicro® X9 5x GPU SuperWorkstation Delivers Maximum Performance with NVIDIA Maximus Certification - May 30th, 2012 [May 30th, 2012]
- Super-virus Flame raises the cyberwar stakes - May 30th, 2012 [May 30th, 2012]
- Super-stealthy ‘Flame' computer virus spies on Iran - May 31st, 2012 [May 31st, 2012]
- Super-stealthy ‘Flame' computer virus spies on Iranians - May 31st, 2012 [May 31st, 2012]
- Was flame virus written by gamers? Code similar to apps such as Angry Birds - May 31st, 2012 [May 31st, 2012]
- Massive cyber attack on Iran came from U.S., report says - June 2nd, 2012 [June 2nd, 2012]
- Massive cyber attack on Iran came from US, report says - June 2nd, 2012 [June 2nd, 2012]
- Supermicro® Exhibits its Latest X9 Server and Storage Innovations at Computex, Taiwan - June 5th, 2012 [June 5th, 2012]
- Supermicro® Hadoop Solutions Accelerate Innovation with Launch of EMC® ... - June 5th, 2012 [June 5th, 2012]
- Super 57000 Video Game (Family Computer) - Video - June 5th, 2012 [June 5th, 2012]
- Security Cameras Turn into Super-Fast Sleuths - June 7th, 2012 [June 7th, 2012]
- Quantum computers move closer to reality, thanks to highly enriched and highly purified silicon - June 7th, 2012 [June 7th, 2012]
- Research Makes Ultrafast Quantum Computer Concept a Reality - June 9th, 2012 [June 9th, 2012]
- Supermicro's New Compact Embedded Server Appliance Supports 3rd Generation Intel® Core™ i7/i5/i3 Processors - June 11th, 2012 [June 11th, 2012]
- The PC which is truly personal: 'Computer' on a memory stick offers COMPLETE privacy for browsing and documents - June 11th, 2012 [June 11th, 2012]
- 'Purified' silicon nudges quantum computing ahead - June 11th, 2012 [June 11th, 2012]
- Apple serves up 15.4-inch MacBook Pro with Retina Display - June 11th, 2012 [June 11th, 2012]
- Apple debuts next-gen MacBook Pro, iOS 6 - June 11th, 2012 [June 11th, 2012]
- How to Invest Like the Super-Rich - June 13th, 2012 [June 13th, 2012]
- Super Computer for Sale - Video - June 13th, 2012 [June 13th, 2012]
- Supermicro® Launches FatTwin™ Architecture - June 15th, 2012 [June 15th, 2012]
- Computer Workstation utilizes NVIDIA® Maximus(TM) technology. - June 15th, 2012 [June 15th, 2012]
- Supermicro® Launches FatTwinâ„¢ Architecture - June 15th, 2012 [June 15th, 2012]
- Acer: Aspire S5, super-thin Ultrabook, coming to U.S. in late June - June 15th, 2012 [June 15th, 2012]
- Supermicro(R) Launches FatTwin(TM) Architecture - June 15th, 2012 [June 15th, 2012]
- Sheldon Adelson: 7 surprising facts about 2012's biggest donor - June 15th, 2012 [June 15th, 2012]
- lego super computer - Video - June 17th, 2012 [June 17th, 2012]
- Age of Empires: The Conqurors - vsing Duke AI 1.6 - Super computer - Video - June 17th, 2012 [June 17th, 2012]
- Supermicro® FatTwin™ Takes Center Stage at International Supercomputing Conference 2012 - June 18th, 2012 [June 18th, 2012]