Jialin Liu , Bradly Crysler - PowerPoint Presentation

Slide1

Jialin Liu, Bradly Crysler, Yin Lu, Yong ChenOct. 07. 2013Data-Intensive Scalable Computing Laboratory (DISCL)

Locality-driven High-level I/O Aggregation for Processing Scientific Datasets

1

OutlineIntroductionMotivation Hila: High Level I/O A

ggregationEvaluationConclusion and Future Work

2

Introduction

Scientific simulations

nowadays generate

a few

terabytes (TB) of data in

a single run

and the data sizes

are expected

to reach

petabytes (PB) in the near future. GCRM, 100 million collumns, 128 levels per column, 50 kmAccessing and analyzing the data reveals poor I/O performance due to the logical-physical mismatching.

IntroductionScientific Datasets and Scientific I/O Libraries

PnetCDF, HDF5, ADIOSPnetCDF

MPI-IO

Parallel File Systems

Scientific I/O libraries allow users to specify array-based logical input

Logical-physical mismatching

Motivation

I/O methods in scientific I/O libraries(

PnetCDF

, ADIOS, HDF5):

Independent I/O

Collective I/O

Noblocking

I/O

Processes collaboration:

No

Calls collaboration :

No

Processes collaboration:

Yes

Calls collaboration :

No

Processes collaboration:

Yes

Calls collaboration :

Yes

MotivationContention on Storage Server without Aware of Locality

…

Call

0

…

Call

1

…

Call

i

…

Two Phase Collective I/O

…

ag

00

ag

01

ag

02

ag

03

…

…

…

ag

1

0

ag

1

1

ag

1

2

ag

1

3

ag

i

0

ag

i

1

ag

i

2

ag

i

3

Performance with Overlapping CallsConclusion: Overlapping Should be Removed

Idea: High level I/O Aggregation

start{0,0,0}

length{100,200,100}

start{0,0,100}

length{100,200,100}

start{10,20,100}

length{10,150,400}

start{10,170,100}

length{10,150,400}

Physical

Layout

sub

0

sub

2

sub

0

sub

2

sub

1

sub

3

sub

1

sub

3

Physical

Layout

start{0,0,0}

length{100,200,200}

start{10,20,100}

length{10,300,400}

Call

0

Call

1

Logical Input

Decomposition

Idea: High level I/O Aggregation

Basic IdeaFigure out the overlapping among requestsEliminate the overlapping before doing I/O

Challenges

How to decompose the requests

How to aggregate the sub-arrays at a high level

Hila: High Level I/O Aggregation

Way to figure out the physical layoutSub-correlation Function

Sub-correlation Set

Lustre Striping: stripe size: s; stripe count: l;

Dataset : Dimension: d; subsets size: m

Hila Algorithm: Prior Step

Prior Step: calculate sub-correlation set, one time analysis

Hila Algorithm: DecompositionMain Steps: Request Decomposition and Aggregation

Improvement with HilaPerformance Improved with Hila

Improvement with HilaFASM Improved with Hila

Conclusion and Future WorkConclusion

The mismatching between logical access and physical layout can lead to poor performance.We propose the locality-driven high-level aggregation approach (HiLa) to facilitate the existing I

/O methods by eliminating the overlapping among sub-array requests.

Future Work

Apply to write operations

Integrate with

f

ile systems.

Locality-driven High-level I/O Aggregationfor Processing Scientific DatasetsThanks

Q&Ahttp://discl.cs.ttu.edu