CyberShake Computational Estimates - Revision history

Scottcal at 15:52, 20 June 2014

2014-06-20T15:52:38Z

Scottcal at 20:56, 18 June 2014

2014-06-18T20:56:59Z

Scottcal at 19:54, 13 May 2014

2014-05-13T19:54:09Z

Scottcal at 21:48, 5 March 2014

2014-03-05T21:48:57Z

Scottcal: /* 1.0 Hz */

2014-03-05T21:46:19Z

1.0 Hz

Scottcal at 20:26, 5 March 2014

2014-03-05T20:26:30Z

Scottcal: /* UCERF 3 */

2013-06-28T14:48:05Z

UCERF 3

Scottcal at 14:47, 28 June 2013

2013-06-28T14:47:50Z

Show changes

Scottcal at 20:48, 14 June 2013

2013-06-14T20:48:17Z

Scottcal at 23:10, 3 April 2012

2012-04-03T23:10:35Z

Show changes

← Older revision		Revision as of 15:52, 20 June 2014
Line 4:		Line 4:

	The 0.5 Hz numbers are taken from Study 14.2.		The 0.5 Hz numbers are taken from Study 14.2.
		+
		+	The node-hours are estimates based on the XE6 and XK7 nodes on Blue Waters.

	== 1.0 Hz, 3 component ==		== 1.0 Hz, 3 component ==

@@ Line 10: / Line 10: @@
    (38 GPU node-hrs per component) x (3 components) x (8 times the gridpoints) x (2 times the timesteps) x (20% more efficient due to more work per GPU) = 1460 node-hrs per site.
-% of node hours to SGTs.
+% of node hours to calculate SGTs.
 PP: At 0.5 Hz, it requires 41 CPU node-hrs per component.
    (41 CPU node-hrs per components) x (3 components) x (25 times the rupture points) x (2 times the timesteps) x (20% more efficient due to rupture generator improvements) = 4920 node-hrs per site.
-% of node hours to PP.
+% of node hours to calculate PP.
 '''6380''' node-hours per 3-component site (181k core-hours)

@@ Line 3: / Line 3: @@
 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.
-== 1.0 Hz ==
+The 0.5 Hz numbers are taken from Study 14.2.
-SGTs: At 0.5 Hz, it requires 35 GPU node-hrs per component.
+== 1.0 Hz, 3 component ==
-PP: At 0.5 Hz, it requires 30 CPU node-hrs per component.
+SGTs: At 0.5 Hz, it requires 38 GPU node-hrs per component.
-   (30 CPU node-hrs per components) x (3 components) x (25 times the rupture points) x (2 times the timesteps) = 4500 node-hrs per site.
+   (38 GPU node-hrs per component) x (3 components) x (8 times the gridpoints) x (2 times the timesteps) x (20% more efficient due to more work per GPU) = 1460 node-hrs per site.
-'''5850''' node-hours per 3-component site (187k core-hours)
+% of node hours to SGTs.
-'''1.67M''' node-hours for standard 3-component So Cal 286-site map (53.5M core-hours)
+PP: At 0.5 Hz, it requires 41 CPU node-hrs per component.
-'''5.25M''' node-hours for increased density 3-component So Cal 898-site map (168M core-hours)
+% of node hours to PP.
-'''8.18M''' node-hours for statewide adaptive 3-component California 1400-site map (262M core-hours)
+'''6380''' node-hours per 3-component site (181k core-hours)
 == 2.0 Hz ==
-SGTs: At 1.0 Hz, it requires 450 GPU node-hrs per component.
+SGTs: At 1.0 Hz, it requires 485 GPU node-hrs per component.
-   (450 GPU node-hrs per component) x (3 components) x (8 times the gridpoints) x (2 times the timesteps) = 22k node-hrs per site.
+   (485 GPU node-hrs per component) x (3 components) x (8 times the gridpoints) x (2 times the timesteps) = 23.3k node-hrs per site.
-PP: At 1.0 Hz, it requires 1500 CPU node-hrs per component.
+PP: At 1.0 Hz, it requires 1640 CPU node-hrs per component.
-   (1500 CPU node-hrs per components) x (3 components) x (2 times the timesteps) = 9k node-hrs per site.
+   (1640 CPU node-hrs per components) x (3 components) x (2 times the timesteps) = 9.8k node-hrs per site.
-'''31k''' node-hours per 3-component site (976k core-hours)
+'''33.1k''' node-hours per 3-component site (686k core-hours)
-'''8.72M''' node-hours for standard 3-component So Cal 286-site map (279M core-hours)
+'''9.47M''' node-hours for standard 3-component So Cal 286-site map (196M core-hours)
-'''108M''' node-hours for increased density 3-component So Cal 3545-site map (3.5B core-hours)
+'''117M''' node-hours for increased density 3-component So Cal 3545-site map (2.4B core-hours)
-'''35M''' node-hours for statewide adaptive 3-component California 1400-site map (1.1B core-hours)
+'''46.3M''' node-hours for statewide adaptive 3-component California 1400-site map (960M core-hours)

@@ Line 6: / Line 6: @@
 SGTs: At 0.5 Hz, it requires 35 GPU node-hrs per component.
-   (35 GPU node-hrs per component) x (3 components) x (8 times the gridpoints) x (2 times the timesteps) x (20% more efficient due to more work per GPU) = 1340 node-hrs per site.
+   (35 GPU node-hrs per component) x (3 components) x (8 times the gridpoints) x (2 times the timesteps) x (20% more efficient due to more work per GPU) = 1350 node-hrs per site.
 PP: At 0.5 Hz, it requires 30 CPU node-hrs per component.
    (30 CPU node-hrs per components) x (3 components) x (25 times the rupture points) x (2 times the timesteps) = 4500 node-hrs per site.
-'''5840''' node-hours per 3-component site (187k core-hours)
+'''5850''' node-hours per 3-component site (187k core-hours)
 '''1.67M''' node-hours for standard 3-component So Cal 286-site map (53.5M core-hours)
@@ Line 21: / Line 21: @@
 == 2.0 Hz ==
-SGTs: At 1.0 Hz, it requires 560 GPU node-hrs per component.
+SGTs: At 1.0 Hz, it requires 450 GPU node-hrs per component.
-   (560 GPU node-hrs per component) x (3 components) x (8 times the gridpoints) x (2 times the timesteps) = 27k node-hrs per site.
+   (450 GPU node-hrs per component) x (3 components) x (8 times the gridpoints) x (2 times the timesteps) = 22k node-hrs per site.
 PP: At 1.0 Hz, it requires 1500 CPU node-hrs per component.
    (1500 CPU node-hrs per components) x (3 components) x (2 times the timesteps) = 9k node-hrs per site.
-'''36k''' node-hours per 3-component site (1.1M core-hours)
+'''31k''' node-hours per 3-component site (976k core-hours)
-'''10.3M''' node-hours for standard 3-component So Cal 286-site map (328M core-hours)
+'''8.72M''' node-hours for standard 3-component So Cal 286-site map (279M core-hours)
-'''127M''' node-hours for increased density 3-component So Cal 3545-site map (4.1B core-hours)
+'''108M''' node-hours for increased density 3-component So Cal 3545-site map (3.5B core-hours)
-'''50.2M''' node-hours for statewide adaptive 3-component California 1400-site map (1.6B core-hours)
+'''35M''' node-hours for statewide adaptive 3-component California 1400-site map (1.1B core-hours)

@@ Line 6: / Line 6: @@
 SGTs: At 0.5 Hz, it requires 35 GPU node-hrs per component.
-   (35 GPU node-hrs per component) x (3 components) x (8 times the gridpoints) x (2 times the timesteps) = 1680 node-hrs per site.
+   (35 GPU node-hrs per component) x (3 components) x (8 times the gridpoints) x (2 times the timesteps) x (20% more efficient due to more work per GPU) = 1340 node-hrs per site.
 PP: At 0.5 Hz, it requires 30 CPU node-hrs per component.
    (30 CPU node-hrs per components) x (3 components) x (25 times the rupture points) x (2 times the timesteps) = 4500 node-hrs per site.
-'''6180''' node-hours per 3-component site (200k core-hours)
+'''5840''' node-hours per 3-component site (187k core-hours)
-'''1.77M''' node-hours for standard 3-component So Cal 286-site map (56.6M core-hours)
+'''1.67M''' node-hours for standard 3-component So Cal 286-site map (53.5M core-hours)
-'''5.55M''' node-hours for increased density 3-component So Cal 898-site map (178M core-hours)
+'''5.25M''' node-hours for increased density 3-component So Cal 898-site map (168M core-hours)
-'''8.65M''' node-hours for statewide adaptive 3-component California 1400-site map (277M core-hours)
+'''8.18M''' node-hours for statewide adaptive 3-component California 1400-site map (262M core-hours)
 == 2.0 Hz ==

@@ Line 69: / Line 69: @@
 == UCERF 3 ==
-For UCERF 3 we are expecting about 350000 ruptures statewide, or 163000 ruptures / 9.7 million rupture variations
+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 ===

@@ 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.
-All these estimates are for UCERF 3, estimating 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 15000 to 200,000, but the number of rupture variations per rupture remains the same.
-== 0.5 Hz (per site) ==
+== Future SCEC data needs ==
 Number of rupture variations:  5.5 million
@@ Line 24: / Line 32: @@
-== 1 Hz (per site) ==
+== 1 Hz, UCERF 3 (per site) ==
 Number of rupture variations:  5.5 million