Anton (computer)

{{Short description|Supercomputer designed and built by D. E. Shaw Research}}

{{Use mdy dates|date=May 2012}}

File:Anton supercomputer.jpg

Anton is a massively parallel supercomputer designed and built by D. E. Shaw Research in New York, first running in 2008. It is a special-purpose system for molecular dynamics (MD) simulations of proteins and other biological macromolecules. An Anton machine consists of a substantial number of application-specific integrated circuits (ASICs), interconnected by a specialized high-speed, three-dimensional torus network.{{cite book

| publisher=ACM

|author1=David E. Shaw |author2=Martin M. Deneroff |author3=Ron O. Dror |author4=Jeffrey S. Kuskin |author5=Richard H. Larson |author6=John K. Salmon |author7=Cliff Young |author8=Brannon Batson |author9=Kevin J. Bowers |author10=Jack C. Chao |author11=Michael P. Eastwood |author12=Joseph Gagliardo |author13=J.P. Grossman |author14=C. Richard Ho |author15=Douglas J. Ierardi |author16=István Kolossváry |author17=John L. Klepeis |author18=Timothy Layman |author19=Christine McLeavey |author20=Mark A. Moraes |author21=Rolf Mueller |author22=Edward C. Priest |author23=Yibing Shan |author24=Jochen Spengler |author25=Michael Theobald |author26=Brian Towles |author27=Stanley C. Wang |title=Anton, a special-purpose machine for molecular dynamics simulation | url=http://cacm.acm.org/magazines/2008/7/5372-anton-a-special-purpose-machine-for-molecular-dynamics-simulation/fulltext

| isbn=978-1-59593-706-3

| journal=Communications of the ACM | volume=51 | issue=7

| date=July 2008 | pages=91–97 | doi=10.1145/1364782.1364802|s2cid=52827083 }} (Related paper published in Proceedings of the 34th Annual International Symposium on Computer Architecture (ISCA '07), San Diego, California, June 9–13, 2007).

Unlike earlier special-purpose systems for MD simulations, such as MDGRAPE-3 developed by RIKEN in Japan, Anton runs its computations entirely on specialized ASICs, instead of dividing the computation between specialized ASICs and general-purpose host processors.

Each Anton ASIC contains two computational subsystems. Most of the calculation of electrostatic and van der Waals forces is performed by the high-throughput interaction subsystem (HTIS).

{{cite book |publisher=IEEE |author1=Richard H. Larson |author2=John K. Salmon |author3=Ron O. Dror |author4=Martin M. Deneroff |author5=Cliff Young |author6=J.P. Grossman |author7=Yibing Shan |author8=John L. Klepeis |author9=David E. Shaw |title=High-Throughput Pairwise Point Interactions in Anton, a Specialized Machine for Molecular Dynamics Simulation |url=http://www.cs.utah.edu/hpca08/papers/6A_1_Larson.pdf |isbn=978-1-4244-2070-4 |journal=Proceedings of the 14th Annual International Symposium on High-Performance Computer Architecture (HPCA '08), Salt Lake City, Utah, February 16–20, 2008 |date=2009 |access-date=January 13, 2009 |archive-url=https://web.archive.org/web/20110605082041/http://www.cs.utah.edu/hpca08/papers/6A_1_Larson.pdf |archive-date=June 5, 2011 |url-status=dead }}

This subsystem contains 32 deeply pipelined modules running at 800 MHz arranged much like a systolic array. The remaining calculations, including the bond forces and the fast Fourier transforms (used for long-range electrostatics), are performed by the flexible subsystem. This subsystem contains four general-purpose Tensilica cores (each with cache and scratchpad memory) and eight specialized but programmable SIMD cores called geometry cores. The flexible subsystem runs at 400 MHz.

{{cite book |publisher=IEEE |author1=Jeffrey S. Kuskin |author2=Cliff Young |author3=J.P. Grossman |author4=Brannon Batson |author5=Martin M. Deneroff |author6=Ron O. Dror |author7=David E. Shaw |title=Incorporating Flexibility in Anton, a Specialized Machine for Molecular Dynamics Simulation |url=http://www.cs.utah.edu/hpca08/papers/6A_2_Kuskin.pdf |isbn=978-1-4244-2070-4 |journal=Proceedings of the 14th Annual International Symposium on High-Performance Computer Architecture (HPCA '08), Salt Lake City, Utah, February 16–20, 2008 |date=2009 |access-date=January 13, 2009 |archive-url=https://web.archive.org/web/20081204154251/http://www.cs.utah.edu/hpca08/papers/6A_2_Kuskin.pdf |archive-date=December 4, 2008 |url-status=dead }}

Anton's network is a 3D torus and thus each chip has 6 inter-node links with a total in+out bandwidth of 607.2 Gbit/s. An inter-node link is composed of two equal one-way links (one traveling in each direction), with each one-way link having 50.6 Gbit/s of bandwidth. Each one-way link is composed of 11 lanes, where a lane is a differential pair of wires signaling at 4.6 Gbit/s. The per-hop latency in Anton's network is 50 ns. Each ASIC is also attached to its own DRAM bank, enabling large simulations.{{cite book

| author = Cliff Young

| author2 = Ron O. Dror

| author3 = J. P. Grossman

| author4 = John K. Salmon

| author5 = Shaw, David E.

| author6 = Joseph A. Bank

| title = Proceedings of the Conference on High Performance Computing Networking, Storage and Analysis

| chapter = A 32x32x32, spatially distributed 3D FFT in four microseconds on Anton

| display-authors = etal

| name-list-style = amp

| year = 2009 |isbn= 978-1-60558-744-8

| pages= 1–11

| publisher = ACM

| location = New York, NY

| doi = 10.1145/1654059.1654083| s2cid = 5611246

}}

The performance of a 512-node Anton machine is over 17,000 nanoseconds of simulated time per day for a protein-water system consisting of 23,558 atoms.

{{cite web|url= http://www.nrbsc.org/anton_rfp/|title= National Resource for Biomedical Supercomputing|access-date= May 14, 2010|archive-url= https://web.archive.org/web/20100523092540/http://www.nrbsc.org/anton_rfp/|archive-date= May 23, 2010|url-status= dead}}

In comparison, MD codes running on general-purpose parallel computers with hundreds or thousands of processor cores achieve simulation rates of up to a few hundred nanoseconds per day on the same chemical system. The first 512-node Anton machine became operational in October 2008.{{cite book |author1=David E. Shaw |author2=Ron O. Dror |author3=John K. Salmon |author4=J.P. Grossman |author5=Kenneth M. Mackenzie |author6=Joseph A. Bank |author7=Cliff Young |author8=Martin M. Deneroff |author9=Brannon Batson |author10=Kevin J. Bowers |author11=Edmond Chow |author12=Michael P. Eastwood |author13=Douglas J. Ierardi |author14=John L. Klepeis |author15=Jeffrey S. Kuskin |author16=Richard H. Larson |author17=Kresten Lindorff-Larsen |author18=Paul Maragakis |author19=Mark A. Moraes |author20=Stefano Piana |author21=Yibing Shan |author22=Brian Towles |chapter=Millisecond-scale molecular dynamics simulations on Anton |title=Proceedings of the Conference on High Performance Computing Networking, Storage and Analysis - SC '09 |year=2009 |isbn=978-1-60558-744-8 |pages=1–11 |format=Portland, Oregon |url=http://cacs.usc.edu/education/cs653/Shaw-msMD-SC09.pdf |publisher=ACM |location=New York, NY, USA |doi=10.1145/1654059.1654099 |s2cid=4390504 |access-date=April 20, 2012 |archive-date=April 23, 2012 |archive-url=https://web.archive.org/web/20120423063209/http://cacs.usc.edu/education/cs653/Shaw-msMD-SC09.pdf |url-status=dead }}

The multiple petaFLOP,{{cite web | url = http://fah-web.stanford.edu/cgi-bin/main.py?qtype=osstats | title = Client Statistics by OS | author = Pande Group | date = Mar 2017 | publisher = Stanford University | access-date =February 3, 2012}} distributed-computing project Folding@home has achieved similar aggregate ensemble simulation timescales, comparable to the total time of a single continuous simulation on Anton, specifically achieving the 1.5-millisecond range in January 2010.{{cite web | url = http://folding.typepad.com/news/2010/01/major-new-result-from-foldinghome-simulation-of-the-millisecond-timescale.html | title = Folding@home: Paper #72: Major new result for Folding@home: Simulation of the millisecond timescale | author = Vijay Pande | date = January 17, 2010 | access-date =September 22, 2011}}

The Anton supercomputer is named after Anton van Leeuwenhoek,{{cite news

| work=The New York Times | author=John Markoff

| title=Herculean Device for Molecular Mysteries

| url=https://www.nytimes.com/2008/07/08/science/08comp.html

| date=July 8, 2008 | access-date=April 25, 2010

}} who is often referred to as "the father of microscopy" because he built high-precision optical instruments and used them to visualize a wide variety of organisms and cell types for the first time.

The ANTON 2 machine with four 512 nodes and its substantially increased speed and problem size has been described.{{cite book

|author1=Shaw, David E |author2=Grossman, JP |author3=Bank, Joseph |author4=A Batson, Brannon |author5=Butts, J Adam |author6=Chao, Jack C |author7=Deneroff, Martin M |author8=Dror, Ron O |author9=Even, Amos |title=SC14: International Conference for High Performance Computing, Networking, Storage and Analysis |chapter=Anton 2: Raising the Bar for Performance and Programmability in a Special-Purpose Molecular Dynamics Supercomputer | year = 2014

| isbn = 978-1-4799-5499-5

| pages = 41–53

| publisher = ACM

| location = New Orleans, LA

| doi = 10.1109/SC.2014.9|s2cid=3354876 }}

The National Institutes of Health have supported an ANTON for the biomedical research community at the Pittsburgh Supercomputing Center, Carnegie Mellon University, and currently (8/20) continues with an ANTON 2 system.