Reducing Checkpoint Size in PlasComCM with Lossy Compression 14th - - PowerPoint PPT Presentation
Reducing Checkpoint Size in PlasComCM with Lossy Compression 14th Annual Workshop on Charm++ and its Applications Jon Calhoun 1 , Franck Cappello 1 , 2 , Luke Olson 1 , and Marc Snir 1 , 2 , Sheng Di 2 1 University of Illinois at Urbana-Champaign
14th Annual Workshop on Charm++ and its Applications
Jon Calhoun1, Franck Cappello1,2, Luke Olson1, and Marc Snir1,2, Sheng Di2
1University of Illinois at Urbana-Champaign 2Argonne National Labatory
19 April 2016
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 1/21
1/21
On current systems, computation is essentially free compared to time and energy required for data transfers. What do we do with these free CPU cycles? [Keckler 2011]
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 2/21
2/21
Native checkpoint restart
Although checkpointing to a partner node is much faster, checkpointing to permanent storage is still needed. Let’s look at improving checkpointing to the parallel filesystem.
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 3/21
3/21
[Son et al. 2014]
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 4/21
4/21
factors [Son et al. 2014]
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 5/21
5/21
High compression ratios with lossy compression [Di and Cappello 2016]
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 6/21
6/21
Whenever you use floating-point values you have already embraced various amounts of error
roundoff.
accurate to the order of the method.
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 7/21
7/21
Many lossy compression schemes allow you to specify an error bound (e.g. relative, absolute).
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 8/21
8/21
Accuracy of numerical methods is expressed as O (hp). Restrict lossy compression error tolerance to be less than truncation error, then error added by lossy compression is hidden in the simulation. Let’s first look at a 1-D heat and a 1-D advection equation to understand what happend to simple PDEs. Setup:
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 9/21
9/21
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 10/21
10/21
0.0 0.5 1.0 1.5 2.0 Time (T) 10
10
10
10
10
10
10
10
Error
Maximum Error at Each Time-step Discretization Error Error due to lossy checkpoint
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 11/21
11/21
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 12/21
12/21
1 2 3 4 5 Time (T) 10
10
10
Error
Maximum Error at Each Time-step of 1-D Advection Discretization Error Error due to lossy checkpoint
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 13/21
13/21
PlasComCM
Setup:
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 14/21
14/21
10
10
10
10
10
Compression Error Tolerance
10 20 30 40 50 60 70 80 90
Compression Factor density energy x momenta y momenta
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 15/21
15/21
10
10
10
10
10
Compression Error Tolerance
0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018
Time (sec) density energy x momenta y momenta
Solid line compression time. Dotted line decompression time.
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 16/21
16/21
Simulation Error
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 17/21
17/21
Simulation Lossy Compressed Simulation
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 18/21
18/21
Simulation Lossy Compressed Simulation
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 18/21
18/21
Lossy compression can effectively reduce the size of a checkpoint without affecting the negatively solution Currently only applicable to file system checkpoints Need to discuss with users to determine acceptable error tolerance Investigate other applications and inputs to gain further insight Further leverage application properties when compressing
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 19/21
19/21
Research under grant FA9550-12-1-0478.
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 20/21
20/21
Jon Calhoun jccalho2@illinois.edu Reducing Checkpoint Size in PlasComCM with Lossy Compression 21/21
21/21