Difference between revisions of "CyberShake Computational Estimates"

From SCECpedia
Jump to navigationJump to search
Line 1: Line 1:
 
We will describe or current best estimates for the CyberShake computational and data requirements as we progress in our simulation planning and testing. These estimates will help us identify which aspects of the CyberShake computational system needs to be optimized to work within our time and resource constraints.
 
We will describe or current best estimates for the CyberShake computational and data requirements as we progress in our simulation planning and testing. These estimates will help us identify which aspects of the CyberShake computational system needs to be optimized to work within our time and resource constraints.
  
−
The UCERF 3 estimates assume that the number of ruptures increases from 15000 to 200,000, but the number of rupture variations per rupture remains the same.
+
The UCERF 3 estimates assume that the number of ruptures increases from 15,000 to 350,000, but the number of rupture variations per rupture on average remains the same.
  
 
== Future SCEC data needs ==
 
== Future SCEC data needs ==
Line 11: Line 11:
 
Spring 2014:  45 TB (0.5 Hz, 286 sites, 1 combination, UCERF 3)
 
Spring 2014:  45 TB (0.5 Hz, 286 sites, 1 combination, UCERF 3)
  
−
== 0.5 Hz, UCERF 3 (per site) ==
+
== UCERF 2 ==
  
−
Number of rupture variations:  5.5 million
+
For UCERF 2 we have about 15,000 ruptures statewide, or 7000 ruptures / 415000 rupture variations per site.
  
−
=== Deterministic ===
+
=== 0.5 Hz (per site), 2 component ===
  
−
Number of jobs:  5.6 million
+
SGT data size:  40 GB
  
−
Storage: 40 GB SGTs, 125 GB seismograms
+
SGT SUs (CPU): 8k
  
−
SUs: 12k SGTs + 26k post-processing = 38k
+
PP number of tasks: 415k
  
−
=== Broadband ===
+
PP SUs: 1k
  
−
Number of jobs: 16.6 million
+
Seismogram data size: 11 GB
  
−
Storage: 40 GB SGTs, 500 GB seismograms
+
Total SUs: 9k
  
−
SUs: 12k SGTs + 52k post-processing = 64k
+
Total data: 51 GB
  
 +
=== 1 Hz (per site), 3 component ===
  
−
== 1 Hz, UCERF 3 (per site) ==
+
SGT data size: 1.9 TB (large increase due to resampling fault surfaces at 250 m)
  
−
Number of rupture variations: 5.5 million
+
SGT SUs (CPU): 192k
  
−
=== Deterministic ===
+
SGT SUs (GPU): 27k
  
−
Number of jobs: 5.6 million
+
PP number of tasks: 415k
  
−
Storage: 320 GB SGTs, 250 GB seismograms
+
PP SUs: 48k
  
−
SUs: 80k SGTs + 150k post-processing = 230k
+
Seismogram data size: 33 GB
  
−
=== Broadband ===
+
Total SUs: 240k CPU / 75k GPU
  
−
Number of jobs: 16.6 million
+
Total data: 1.9 TB
  
−
Storage:  320 GB SGTs, 2 TB seismograms
+
=== 1.5 Hz (per site), 3 component ===
  
−
SUs:  80k SGTs + 200k post-processing = 280k
+
SGT data size:  2.8 TB
  
−
== Southern California simulations ==
+
SGT SUs (CPU): 648k
  
−
200 sites
+
SGT SUs (GPU): 93k
−
=== 0.5 Hz, deterministic ===
 
  
−
Number of jobs: 1.1 billion
+
PP number of tasks: 415k
  
−
Storage: 7.8 TB SGTs, 24.4 TB seismograms
+
PP SUs: 72k
  
−
SUs: 7.6 million
+
Seismogram data size: 50 GB
  
−
=== 0.5 Hz, broadband ===
+
Total SUs: 720k CPU / 165k GPU
  
−
Number of jobs: 3.3 billion
+
Total data: 2.8 TB
  
−
Storage:  7.8 TB SGTs, 97 TB seismograms
+
== UCERF 3 ==
  
−
SUs: 12.8 million
+
For UCERF 3 we are expecting about 350000 ruptures statewide, or 163000 ruptures / 9.7 million rupture variations
  
−
=== 1 Hz, deterministic ===
+
=== 0.5 Hz (per site), 2 component ===
  
−
Number of jobs:  1.1 billion
+
SGT data size:  930 GB
  
−
Storage: 62.5 TB SGTs, 48.8 TB seismograms
+
SGT SUs (CPU): 8k
  
−
SUs: 46 million
+
PP number of tasks: 9.7M
  
−
=== 1 Hz, broadband ===
+
PP SUs: 23k
  
−
Number of jobs: 3.3 billion
+
Seismogram data size: 256 GB
  
−
Storage: 62.5 TB SGTs, 400 TB seismograms
+
Total SUs: 31k
  
−
SUs: 56 million
+
Total data: 1.1 TB
  
−
== Statewide simulations ==
+
=== 1 Hz (per site), 3 component ===
  
−
1400 sites
+
SGT data size: 44.5 TB
−
=== 0.5 Hz, deterministic ===
 
  
−
Number of jobs: 7.8 billion
+
SGT SUs (CPU): 192k
  
−
Storage: 54.7 TB SGTs, 171 TB seismograms
+
SGT SUs (GPU): 27k
  
−
SUs: 53 million
+
PP number of tasks: 9.7M
  
−
=== 0.5 Hz, broadband ===
+
PP SUs: 1.1M
  
−
Number of jobs: 23.2 billion
+
Seismogram data size: 768 GB
  
−
Storage: 54.7 TB SGTs, 683 TB seismograms
+
Total SUs: 240k CPU / 75k GPU
  
−
SUs: 89.6 million
+
Total data: 45.2 TB
  
−
=== 1 Hz, deterministic ===
+
=== 1.5 Hz (per site), 3 component ===
  
−
Number of jobs:  7.8 billion
+
SGT data size:  65.4 TB
  
−
Storage: 437.5 TB SGTs, 341.8 TB seismograms
+
SGT SUs (CPU): 648k
  
−
SUs: 322 million
+
SGT SUs (GPU): 93k
  
−
=== 1 Hz, broadband ===
+
PP number of tasks: 9.7M
  
−
Number of jobs: 23.2 billion
+
PP SUs: 1.7M
  
−
Storage: 437.5 TB SGTs, 2800 TB seismograms
+
Seismogram data size: 1.1 TB
  
−
SUs: 392 million
+
Total SUs: 720k CPU / 165k GPU
 +
 
 +
Total data: 66.5 TB

Revision as of 14:47, 28 June 2013

We will describe or current best estimates for the CyberShake computational and data requirements as we progress in our simulation planning and testing. These estimates will help us identify which aspects of the CyberShake computational system needs to be optimized to work within our time and resource constraints.

The UCERF 3 estimates assume that the number of ruptures increases from 15,000 to 350,000, but the number of rupture variations per rupture on average remains the same.

Future SCEC data needs

These are estimates of CyberShake storage required on SCEC computers for upcoming runs.

July-August 2013: 6.1 TB to archive (0.5 Hz, CVM-SI, 2 SGT versions, 286 sites, UCERF 2) Fall 2013: 12 TB (1 Hz, 286 sites, 2 combinations, UCERF 2) Spring 2014: 45 TB (0.5 Hz, 286 sites, 1 combination, UCERF 3)

UCERF 2

For UCERF 2 we have about 15,000 ruptures statewide, or 7000 ruptures / 415000 rupture variations per site.

0.5 Hz (per site), 2 component

SGT data size: 40 GB

SGT SUs (CPU): 8k

PP number of tasks: 415k

PP SUs: 1k

Seismogram data size: 11 GB

Total SUs: 9k

Total data: 51 GB

1 Hz (per site), 3 component

SGT data size: 1.9 TB (large increase due to resampling fault surfaces at 250 m)

SGT SUs (CPU): 192k

SGT SUs (GPU): 27k

PP number of tasks: 415k

PP SUs: 48k

Seismogram data size: 33 GB

Total SUs: 240k CPU / 75k GPU

Total data: 1.9 TB

1.5 Hz (per site), 3 component

SGT data size: 2.8 TB

SGT SUs (CPU): 648k

SGT SUs (GPU): 93k

PP number of tasks: 415k

PP SUs: 72k

Seismogram data size: 50 GB

Total SUs: 720k CPU / 165k GPU

Total data: 2.8 TB

UCERF 3

For UCERF 3 we are expecting about 350000 ruptures statewide, or 163000 ruptures / 9.7 million rupture variations

0.5 Hz (per site), 2 component

SGT data size: 930 GB

SGT SUs (CPU): 8k

PP number of tasks: 9.7M

PP SUs: 23k

Seismogram data size: 256 GB

Total SUs: 31k

Total data: 1.1 TB

1 Hz (per site), 3 component

SGT data size: 44.5 TB

SGT SUs (CPU): 192k

SGT SUs (GPU): 27k

PP number of tasks: 9.7M

PP SUs: 1.1M

Seismogram data size: 768 GB

Total SUs: 240k CPU / 75k GPU

Total data: 45.2 TB

1.5 Hz (per site), 3 component

SGT data size: 65.4 TB

SGT SUs (CPU): 648k

SGT SUs (GPU): 93k

PP number of tasks: 9.7M

PP SUs: 1.7M

Seismogram data size: 1.1 TB

Total SUs: 720k CPU / 165k GPU

Total data: 66.5 TB