Clusters, Grids and their applications in Physics David Barnes (Astro) Lyle Winton (EPP)

Clusters, Grids and their applications in Physics

David Barnes (Astro)

Lyle Winton (EPP)

Today’s typical workstation can:• Compute 1024-point fast Fourier transforms at a

rate of 40000 per second.• Compute and apply gravitational force for 32768

particles at a rate of ~one time step per sec.• Render ~7M elements of a data volume per sec.• Stream data to and from disk at around 30 MByte

per sec.• Communicate with other machines at up to 10

Mbyte per sec, with a latency of a few tens of milliseconds.

… … what if this is not enough? … …

High Performance Computers

• ~ 20 years ago– 1x106 Floating Point Ops/sec (Mflop/s) [scalar processors]

• ~ 10 years ago– 1x109 Floating Point Ops/sec (Gflop/s) [vector processors]

• ~ Today: superscalar-based CLUSTERS– 1x1012 Floating Point Ops/sec (Tflop/s)

• Highly parallel, distributed, networked superscalar processors

• ~ less than 10 years away– 1x1015 Floating Point Ops/sec (Pflop/s)

• Multi-level clusters and GRIDS

a+b=c

A+B=C

a+b=cd+e=f

High-Performance Computing Directions: Beowulf-class PC Clusters

• Common off-the-shelf PC Nodes– Pentium, Alpha, PowerPC, SMP

• COTS LAN/SAN Interconnect– Ethernet, Myrinet, Giganet, ATM

• Open Source Unix– Linux, BSD

• Message Passing Computing– MPI, PVM– HPF

• Best price-performance• Low entry-level cost• Just-in-place configuration• Vendor invulnerable• Scalable• Rapid technology tracking

Definition: Advantages:

Enabled by PC hardware, networks and operating system achieving capabilities of scientific workstations at a fraction of the cost and availability of industry standard message passing libraries.

Slide from Jack Dongarra… let’s see some clusters …

Mojo: School of Physics cluster24 nodes, 24 CPUs, 2 & 2.4 GHz Pentium 4

~70 Gflop/s~70 Gflop/s

Swinburne Centre for Astrophysics and Supercomputing: 90 nodes, 180 CPUs, 2.0, 2.2 & 2.4 GHz Pentium 4

16 nodes, 32 CPUs, 933 MHz Pentium III

~500 Gflop/s~500 Gflop/s• Accomplish > 8 million

1024-pt FFTs per second.• Render > 109 volume

elements per second.• Calculate at one time step

per second for ~150000 particles using brute-force approach.

Next-generation “blade” clusterIBM Blue Gene ~200 Tflop/s~200 Tflop/s

SETI@home• Uses thousands of Internet-connected

PCs to help in the search for extraterrestrial intelligence.

• Uses data collected with the Arecibo Radio Telescope, in Puerto Rico

• When your computer is idle the software downloads a 300 kilobyte chunk of data for analysis.

• The results of this analysis are sent back to the SETI team and combined with results from thousands of other participants.

• Largest distributed computation project in existence– ~ 400,000 machines

– Averaging 26 Tflop/s

Slide from Jack Dongarra

This is more than a cluster … perhaps it is the first genuine “Grid” …

http://setiathome.berkeley.edu

over to Lyle Winton…