single virtual systems for life and material sciences
Post on 04-Jun-2018
216 Views
Preview:
TRANSCRIPT
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
1/21
SINGLE VIRTUAL SYSTEMFORLIFE AND MATERIAL SCIENCES
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
2/21
2
Small groups of scientists: experts in their domain,
not in computer science
The more compute power, the more simulationsreflect the real world
Many different types of applications, some parallel,some sequential, some requiring vast amounts ofmemory
Research relies on commercially available
applications as well as in-house development
The faster the simulations, the faster results arepublished, the faster new products hit their market.
KEY ISSUES IN LIFE ANDMATERIAL SCIENCES
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
3/21
3
APPLICATIONS SEGMENTATION
(*) QSPR: Quantitative Structure Property Relationship
CAN ONE COMPUTING ARCHITECTURE
RUN BOTH?
Compute
Memory
andData
Systems Biology Monte Carlo Simulations
Stochastic Dynamics
Genetics
Image processing and Analysis
Sequence Analysis Genomics
Proteomics
Picture Archiving
Docking
Molecular Analysis
QSPR(*)
Lab Data Management
Quantum Chemistry Molecular Dynamics
Molecular Mechanics
Quantum Chemistry Whole Genomics
In Memory Database
Searching
High Throughput Large Simulation
Cluster
Codes
SMPCodes
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
4/21
4
1 Operating System
1.6 TeraFLOPS
128-cores
2.3TB RAM
19TB Internal Storage
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
5/21
5
SCALEMP AT A GLANCE
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
6/21
6
NEW SERVER VIRTUALIZATIONPARADIGMENTERPRISEAPPLICATIONS HIGH PERFORMANCECOMPUTING
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
7/21
7
SERVER VIRTUALIZATIONAGGREGATION
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
8/21
8
VSMP FOUNDATION AGGREGATIONPLATFORM
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
9/21
9
VSMP FOUNDATION AGGREGATIONPLATFORM
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
10/21
10
1
0
VSMP FOUNDATION FOR SMPDIRECT CONNECT 2
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
11/21
11
VALUE PROPOSITIONS
Overcoming the
challenges of
applications and
end users
Overcoming thechallenges of IT
administration
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
12/21
12
HOW IT WORKS
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
13/21
13
BEHIND THE SCENES
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
14/21
14
DELL SUPPORTED HARDWARE
PowerEdge M1000e with
M600/M610/M710 blades
Aggregation of 2 to 16 systems
Highest x86 shared-memory supercomputer
Dell Rack-Mount Servers
PE1950III/R410/R610/R710
Up to 32 Nehalem Processors (128 cores) and 2.3TB of Shared Memory
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
15/21
15
SINGLE VIRTUAL SYSTEM
Typical Applications
Gaussian
VASP
AMBER
Schrdinger Jaguar
Schrdinger Glide
NAMDDOCK
GAMESS
GOLD
mpiBLAST
GROMACS
MOLPRO
OpenEye FRED
OpenEye OMEGASCM ADF
HMMER.
.
.
Single Operating System
One Large Shared Memory
Hidden InfiniBand Infrastructure
Simplified Storage
Simplified Management
MANAGEABLE PERFORMANCE
Focusing on Science not Computer Science
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
16/21
16
EXAMPLE OF A GENERAL PURPOSELMS FAT NODE IMPLEMENTATION
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
17/21
17
COMPUTATIONAL CHEMISTRY
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
18/21
18
MOLECULAR DYNAMICS
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
19/21
19
VSMPROFILE:PERFORMANCE TUNINGCOMPLETE SYSTEM VIEWBefore After
FOR APPLICATIONS DEVELOPERS:
Points to line of code causing performance inefficiencies
Eliminates performance optimization guesswork
Significantly reduces application tuning time
Not available on other architectures
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
20/21
20
Small groups of scientists: experts in their domain,
not in computer science
The more compute power, the more simulationsreflect the real world
Many different types of applications, some parallel,some sequential, some requiring vast amounts ofmemory
Research relies on commercially available
applications as well as in-house development
The faster the simulations, the faster results arepublished, the faster new products hit their market.
KEY ISSUES IN LIFE ANDMATERIAL SCIENCES SOLVED
-
8/13/2019 Single Virtual Systems for Life and Material Sciences
21/21
Thank you
top related