Modeling Resource-Coupled Computations MarkHereld - - PowerPoint PPT Presentation

▶
modeling resource coupled computations
SMART_READER_LITE
LIVE PREVIEW

Modeling Resource-Coupled Computations MarkHereld - - PowerPoint PPT Presentation

Jul 21, 2023 140 likes •404 views

Modeling Resource-Coupled Computations MarkHereld Computa0onIns0tute Mathema0csandComputerScience ArgonneLeadershipCompu0ngFacility ArgonneNa0onalLaboratory UniversityofChicago Roadmap

slide-1
SLIDE 1 Mark
Hereld
 Computa0on
Ins0tute
 Mathema0cs
and
Computer
Science
 Argonne
Leadership
Compu0ng
Facility
 Argonne
Na0onal
Laboratory
 University
of
Chicago


Modeling Resource-Coupled Computations

slide-2
SLIDE 2

Roadmap

  • issues
and
ideas

  • models
and
measurements

  • implica0ons
and
work
in
progress

slide-3
SLIDE 3

Issue

  • Given
increasingly
massive
(and
complex)
datasets…

  • how
to
connect
them
to
computa0onal
and
display

resources
that
support
visualiza0on
and
analysis?



  • holis0c
approaches
to
alloca0ng
simula0on,
analysis,

visualiza0on,
display,
storage,
and
network
resources

  • create
and
exploit
ways
to
op0mally
couple
these

resources
in
real
0me

slide-4
SLIDE 4

Common sense

  • Analysis
engines
must
be
co‐located
with
simula0on

engines

  • …or
even,
analysis
code
must
be
co‐located
with

simula0on
code,
i.e
in
situ

  • Display
resources
must
be
integrated
locally
with
HPC

resources

  • In
general,
wide‐area
applica0ons
will
become

impossible…

  • But,
maybe
the
situa0on
isn’t
so
dire.

slide-5
SLIDE 5

ideas

  • Ideas

  • Models

  • Measurements

  • Consequences

  • Future



ideas


slide-6
SLIDE 6

Mitigation

  • More
efficient
I/O
prac0ces

– Many
(most)
inefficiencies
in
R/W
rates
amenable
to
beWer
 prac0ces
by
applica0on
developer
 – In
addi0on
to
improvements
in
performance
of
I/O
libraries

  • BeWer
data
management

– BeWer
data
layout

  • BeWer
brute
force
compression
methods

– Uncertainty
aware;
domain
aware

  • Leveraging
limita0ons
at
the
des0na0on

– Pixel
real
estate
 – Perceptual
limita0ons
(and
features)

slide-7
SLIDE 7

Coupled Resources

  • remote
visualiza0on:
couple
data
and
large

computa0onal
resources
to
remote
display
hardware

  • in
situ
analysis
and
visualiza0on:
merge
simula0on

and
analysis
code
on
single
machine

  • co‐analysis:
couple
simula0on
on
supercomputer
to

live
analysis
on
visualiza0on
and
analysis
plaZorm

slide-8
SLIDE 8

models

  • Ideas

  • Models

  • Measurements

  • Consequences

  • Future



models


slide-9
SLIDE 9 128
File‐Server
 
Nodes
 Eureka
 100
Nodes
 10GE
 640
x
10G
 
=
6.4
Tbps
 Myrinet
Switch
 
Complex
 5‐Stage
CLOS

 
10GE<‐>MX
conversion
 MX<‐>MX
 640
BGP
I/O
Nodes
 40K
BGP
Compute
Nodes
 10G
MX
 100
x
10G
 
=
1
Tbps
 128
x
10G
 
=
1.28
Tbps
 10G
MX
 Tree
 4.3Tbps
 Theore0cal
Max
Bandwidth
from
I/O
Nodes
to
Eureka
(Memory
to
Memory)



=
1
Tbps

 
 
 
 
 
 
 
 
 
 
 
Bi‐direc0onal 


=
2
Tbps
 Theore0cal
Max
Bandwidth
from
I/O
Nodes
to
FileServer
(Memory
to
Memory)
=
1.28
Tbps
 
 
 
 
 
 
 
 
 
 
 
Bi‐direc0onal 



=
2.56
Tbps
 Theore0cal
Max
Bandwidth
from
Eureka
to
FileServer
(Memory
to
Memory)




=
1
Tbps
 
 
 
 
 
 
 
 
 
 
 
Bi‐direc0onal 



=
2
Tbps
 ALCF
Network
Architecture
 Tbps
–
Terabits/sec

slide-10
SLIDE 10

Data Analytics Resource: Eureka

  • Data
analy0cs
and
visualiza0on
cluster
at
ALCF

  • (2)
head
nodes,
(100)
compute
nodes

– (2)
Nvidia
Quadro
FX5600
graphics
cards
 – (2)
XEON
E5405
2.00
GHz
quad
core
processors
 – 32
GB
RAM:
(8)
4
rank,
4GB
DIMMS
 – (1)
Myricom
10G
CX4
NIC
 – (2)
250GB
local
disks;
(1)
system,
(1)
minimal
scratch
 – 32
GFlops
per
server

slide-11
SLIDE 11

Application

  • FLASH

– Mul0‐physics
code:
Gravita0on,
nuclear
chemistry,
MHD
 – Laboratory
to
Universe

  • Mul0ple
(~20)
simula0ons

– 8km
resolu0on,
10K
to
100K
blocks
each
(16
*
16
*
16)
voxel
 – 2
Racks
(8K
cores)
of
the
ANL’s
Intrepid
(BGP)
 – typical
simula0on
is
10
runs
each
12
hours

  • O(hour)
per
checkpoint
cycle

– 66%
0me
spent
simula0ng
 – 33%
0me
spent
non‐overlapping
I/O

slide-12
SLIDE 12

measurements

  • Ideas

  • Models

  • Measurements

  • Consequences

  • Future



measurements


slide-13
SLIDE 13

Flash IO for 1 run (12 hours)

  • Total
Run
0me

=

41557
secs

– IO
0me
during
run
=
14325
sec

(34%
of
the
0me)
 – Circa
March
2009

  • Par0cle
Data:

– 417
Files
(0.1GB
each)
=

41.7
GB
 – Time
spent
wri0ng
=

9047
secs
(
22%
of
the
run
0me)

  • Plot
files:

– 104
files
(2.5GB
each)
;Total
=
260GB

 – Time
spent
in
wri0ng
=

3897
secs
(
9%
of
the
run
0me)

  • Checkpoint

files:

– 10
files
(8
GB
each)
;Total
=
80GB

 – Time
spent
in
wri0ng
=
1144
secs
(
3%
of
the
run
0me)

slide-14
SLIDE 14

FLASH Supernova Explosion Project

  • mul0ple
(~20)
simula0ons

– 8km
resolu0on
 – 10K
to
100K
blocks
each
(16
*
16
*
16)
voxel
 – 2
Racks
(8K
cores)
of
the
ANL’s
Intrepid
(BGP)
 – typical
simula0on
is
10
runs
each
12
hours
 – Circa
November
2009

  • =======================================================
  • File Type File Size #files #files Data Size
  • / Run / Sim
  • =======================================================
  • Particle
~ 131 MB ~ 500 5000 500 GB
  • Plot ~ 13 GB 40-90 800 10 TB
  • Checkpoint ~ 42 GB 5-10 100 4.2 TB
  • =======================================================
slide-15
SLIDE 15

Internal Network Experiments

Tree
Network
 Switch
 BGP
Compute
Nodes
 BGP
I/O
Node
 Analysis
Node

slide-16
SLIDE 16

Toward middleware to facilitate co-analysis

BGP
Compute
Nodes

slide-17
SLIDE 17

consequences

  • Ideas

  • Models

  • Measurements

  • Consequences

  • Future



consequences


slide-18
SLIDE 18

Map Intrepid I/O to Eureka

  • Speed
up
the
applica0on

– Offload
data
organiza0on
and
disk
writes

  • Free
co‐analysis

– Produce
several
high
resolu0on
movies
 – Data
compression
 – Mul0‐0me
step
caching
for
window
analysis

  • Eureka
is
an
accelerator
and
co‐analysis
engine
at
only

1‐2%
cost
of
Intrepid

slide-19
SLIDE 19

future

  • Ideas

  • Models

  • Measurements

  • Consequences

  • Future



future


slide-20
SLIDE 20

Works in Progress

  • Footprints

– System
level
use
paWern
data
collec0on
 – Boo0ng
up
a
mini‐consor0um
of
resource
monitoring
enthusiasts

  • in
situ

– Papka
parallel
sorware
rendering
 – Tom
Peterka
and
Rob
Ross
scaling
sorware
rendering
algorithms
 – HW‐SW
rendering
comparison
experiments

  • Co‐analysis

– StarGate
experiments
 – Intrepid
<>
Eureka
communica0on
experiments
 – FLASH
test

  • Remote
Visualiza0on

– Pixel
shipping
experiments
and
frameworks

slide-21
SLIDE 21 0.00001 0.0001 0.001 0.01 0.1 1 1 10 100 1000 Time (secs) Num Procs Eureka Rendering Times 256x256x256 Full Frame Time Render Time Composite Network Time Composite Render Time Sync State Time 0.00001 0.0001 0.001 0.01 0.1 1 1 10 100 1000 Time (secs) Num Procs Eureka Rendering Times 512x512x512 Full Frame Time Render Time Composite Network Time Composite Render Time Sync State Time 0.00001 0.0001 0.001 0.01 0.1 1 10 1 10 100 1000 Time (secs) Num Procs Eureka Rendering Times 1024x1024x1024 Full Frame Time Render Time Composite Network Time Composite Render Time Sync State Time 0.0001 0.001 0.01 0.1 1 1 10 100 Time (secs) Num Procs Eureka Rendering Times 2048x2048x2048 Full Frame Time Render Time Composite Network Time Composite Render Time Sync State Time 0.0001 0.001 0.01 0.1 1 10 100 1 10 100 1000 Time (secs) Num Procs Surveyor Rendering Times 256x256x256 Full Frame Time Render Time Composite Network Time Composite Render Time Sync State Time 0.0001 0.001 0.01 0.1 1 10 100 1 10 100 1000 Times (secs) Num Procs Surveyor Rendering Times 512x512x512 Full Frame Time Render Time Composite Network Time Composite Render Time Sync State Time
slide-22
SLIDE 22

Wide Area Experiments

Simula0on

  • 4K
uniform
grid
cube

  • Single
variable,
float

  • 257
GB
per
0me
step

  • 577
0me
steps

  • 150
TB
total

Visualiza0on

  • Volume
rendering

  • 4K
x
4K
pixel

Interac0ve
Display

  • Large
0led
display

  • Naviga0on

  • Manipula0on

RAW
 DATA
 RESULTS
 CONTROL
 DETAILS
AND
DEMO
IN
SDSU
BOOTH

slide-23
SLIDE 23

Summary

  • Discussion
of
the
issues
with
illumina0ng
example

– Presumed
impending
Doom
outlined

  • Discussion
of
the
ideas
with
examples

– Resource‐coupled
computa0ons

  • In
situ
couples
simula0on
and
analysis
in
real
0me
on
shared
compute

  • Remote
vis
couples
compute
and
data
resources
to
remote
display
clients

  • Co‐analysis
couples
two
compute
resources
in
real
0me

  • Discussion
of
the
work
in
progress
with
status

– Suite
of
experiments
underway
to
characterize
system
components
 – Strawman
use
cases
in
place
provide
challenging
and
exci0ng
goals
 – Stunning
results
and
paradigm
shirs
forthcoming

slide-24
SLIDE 24

Acknowledgements

  • Venkat
Vishwanath

  • Michael
Papka

  • Eric
Olson

  • Joe
Insley

  • Tom
Uram

  • Tom
Peterka

  • Rob
Ross

  • Rick
Stevens

  • Rick Wagner, UCSD
  • Michael Norman, UCSD
  • Robert Harkess (UCSD)
  • Narayan Desai
  • David Ressman
  • William Scullin
  • Loren Wilson
  • Linda Winkler
  • ESNET2
slide-25
SLIDE 25

end

  • Ques0ons?


end