Cluster WorkshopCluster Workshop

For COMP RPG students

17 May, 2010

High Performance Cluster Computing Centre (HPCCC)Faculty of Science

Hong Kong Baptist University

2

Outline• Overview of Cluster Hardware and Overview of Cluster Hardware and

SoftwareSoftware• Basic Login and Running Program in a Basic Login and Running Program in a

job queuing systemjob queuing system • Introduction to ParallelismIntroduction to Parallelism

– Why ParallelismWhy Parallelism– Cluster ParallelismCluster Parallelism

• Open MPOpen MP• Message Passing InterfaceMessage Passing Interface• Parallel Program ExamplesParallel Program Examples• Policy Policy for usingfor using sciblade.sci.hkbu.edu.hksciblade.sci.hkbu.edu.hkhttp://www.sci.hkbu.edu.hk/hpccc/sciblade

2

Overview of Cluster Overview of Cluster Hardware and SoftwareHardware and Software

4

Cluster Hardware

This 256-node PC cluster (sciblade) consist of:

• Master node x 2• IO nodes x 3 (storage)• Compute nodes x 256• Blade Chassis x 16• Management network• Interconnect fabric• 1U console & KVM switch• Emerson Liebert Nxa 120k VA UPS

4

5

Sciblade Cluster

256-node clusters supported by fund from RGC

5

6

Hardware Configuration

• Master Node– Dell PE1950, 2x Xeon E5450 3.0GHz (Quad Core)– 16GB RAM, 73GB x 2 SAS drive

• IO nodes (Storage)– Dell PE2950, 2x Xeon E5450 3.0GHz (Quad Core)– 16GB RAM, 73GB x 2 SAS drive– 3TB storage Dell PE MD3000

• Compute nodes x 256 each with– Dell PE M600 blade server w/ Infiniband network – 2x Xeon E5430 2.66GHz (Quad Core)– 16GB RAM, 73GB SAS drive

6

7

Hardware Configuration• Blade Chassis x 16

– Dell PE M1000e – Each hosts 16 blade servers

• Management Network– Dell PowerConnect 6248 (Gigabit Ethernet) x 6

• Inerconnect fabric– Qlogic SilverStorm 9120 switch

• Console and KVM switch – Dell AS-180 KVM– Dell 17FP Rack console

• Emerson Liebert Nxa 120kVA UPS7

8

Software List• Operating System

– ROCKS 5.1 Cluster OS– CentOS 5.3 kernel 2.6.18

• Job Management System – Portable Batch System– MAUI scheduler

• Compilers, Languages – Intel Fortran/C/C++ Compiler for Linux V11– GNU 4.1.2/4.4.0 Fortran/C/C++ Compiler

8

9

Software List

• Message Passing Interface (MPI) Libraries – MVAPICH 1.1– MVAPICH2 1.2– OPEN MPI 1.3.2

• Mathematic libraries – ATLAS 3.8.3– FFTW 2.1.5/3.2.1– SPRNG 2.0a(C/Fortran) /4.0(C++/Fortran)

9

10

Software List

• Molecular Dynamics & Quantum Chemistry– Gromacs 4.0.7

– Gamess 2009R1,

– Gaussian 03

– Namd 2.7b1

• Third-party Applications – FDTD simulation– MATLAB 2008b– TAU 2.18.2, VisIt 1.11.2– Xmgrace 5.1.22

– etc

10

11

Software List

• Queuing system– Torque/PBS– Maui scheduler

• Editors– vi– emacs

11

12

Hostnames

• Master node– External : sciblade.sci.hkbu.edu.hk– Internal : frontend-0

• IO nodes (storage)– pvfs2-io-0-0, pvfs2-io-0-1, pvfs-io-0-2

• Compute nodes– compute-0-0.local, …, compute-0-255.local

12

Basic Login and Running Program in a Job Queuing System

14

Basic login• Remote login to the master node• Terminal login

– using secure shellssh -l username sciblade.sci.hkbu.edu.hk

• Graphical login– PuTTY & vncviewer e.g.

[username@sciblade]$ vncserver

New ‘sciblade.sci.hkbu.edu.hk:3 (username)' desktop is sciblade.sci.hkbu.edu.hk:3

It means that your session will run on display 3.

14

15

Graphical login

• Using PuTTY to setup a secured connection: Host Name=sciblade.sci.hkbu.edu.hk

15

16

Graphical login (con’t)

• ssh protocol version

16

17

Graphical login (con’t)

• Port 5900 +display number (i.e. 3 in this case)

17

18

Graphical login (con’t)• Next, click Open, and login to sciblade• Finally, run VNC Viewer on your PC, and enter

"localhost:3" {3 is the display number}

• You should terminate your VNC session after you have finished your work. To terminate your VNC session running on sciblade, run the command[username@tdgrocks] $ vncserver –kill : 3

18

19

Linux commands• Both master and compute nodes are installed with

Linux• Frequently used Linux command in PC cluster

http://www.sci.hkbu.edu.hk/hpccc/sciblade/faq_sciblade.php

cp cp f1 f2 dir1 copy file f1 and f2 into directory dir1

mv mv f1 dir1 move/rename file f1 into dir1

tar tar xzvf abc.tar.gz Uncompress and untar a tar.gz format file

tar tar czvf abc.tar.gz abc create archive file with gzip compression

cat cat f1 f2 type the content of file f1 and f2

diff diff f1 f2 compare text between two files

grep grep student * search all files with the word student

history history 50 find the last 50 commands stored in the shell

kill kill -9 2036 terminate the process with pid 2036

man man tar displaying the manual page on-line

nohup nohup runmatlab a run matlab (a.m) without hang up after logout

ps ps -ef find out all process run in the systems

sort sort -r -n studno sort studno in reverse numerical order

19

20

ROCKS specific commands

• ROCKS provides the following commands for users to run programs in all compute node. e.g.– cluster-fork

• Run program in all compute nodes

– cluster-fork ps• Check user process in each compute node

– cluster-kill• Kill user process at one time

– tentakel• Similar to cluster-fork but run faster

20

21

GangliaWeb based management and monitoring• http://sciblade.sci.hkbu.edu.hk/ganglia

21

Why ParallelismWhy Parallelism

23

Why Parallelism – Passively

• Suppose you are using the most efficient algorithm with an optimal implementation, but the program still takes too long or does not even fit onto your machine

• Parallelization is the last chance.

23

24

Why Parallelism – Initiative

• Faster– Finish the work earlier

= Same work in shorter time

– Do more work= More work in the same time

• Most importantly, you want to predict the result before the event occurs

24

25

ExamplesMany of the scientific and engineering problems require enormous computational power.

Following are the few fields to mention:

– Quantum chemistry, statistical mechanics, and relativistic physics

– Cosmology and astrophysics

– Computational fluid dynamics and turbulence

– Material design and superconductivity

– Biology, pharmacology, genome sequencing, genetic engineering, protein folding, enzyme activity, and cell modeling

– Medicine, and modeling of human organs and bones

– Global weather and environmental modeling

– Machine Vision

25

26

Parallelism

• The upper bound for the computing power that can be obtained from a single processor is limited by the fastest processor available at any certain time.

• The upper bound for the computing power available can be dramatically increased by integrating a set of processors together.

• Synchronization and exchange of partial results among processors are therefore unavoidable.

26

27

Multiprocessing Clustering

IS

CU CU CU CU

PU PU PU PU

Shared Memory

1 n-1 n2

21 n-1 n

ISISIS

DSDSDSDS

DS

LM LM LM LM

CPU CPU CPU CPU

Interconnecting Network

1 n-1 n2

21 n-1 n

DSDSDS

Distributed Memory – Cluster

Shared Memory – Symmetric multiprocessors (SMP)

Parallel Computer Architecture

27

28

Clustering: Pros and Cons

• Advantages – Memory scalable to number of processors.

∴Increase number of processors, size of memoryand bandwidth as well.

– Each processor can rapidly access its own memory without interference

• Disadvantages – Difficult to map existing data structures to this

memory organization – User is responsible for sending and receiving data

among processors

28

29

TOP500 Supercomputer Sites (www.top500.org)

Architecture of Top 500

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009Year

Sh

are

pe

rce

nta

ge

SMP Constellations

MPP Cluster

29

Cluster ParallelismCluster Parallelism

31

Parallel Programming Paradigm

Multithreading – OpenMP

Message Passing– MPI (Message Passing Interface)– PVM (Parallel Virtual Machine)

Shared memory, Distributed memory

Shared memory only

31

32

Distributed Memory• Programmers view:

– Several CPUs– Several block of memory– Several threads of action

• Parallelization– Done by hand

• Example– MPI

time

P1

P1 P2 P3

P2

P3

Process 0

Process 1

Process 2

SerialSerial

Data exchange viainterconnection

Process

MessageMessagePassingPassing

32

33

Message Passing Model

Message Passing

The method by which data from one processor's memory is copied to the memory of another processor.

ProcessA process is a set of executable instructions (program) which runs on a processor.

Message passing systems generally associate only one process per processor, and the terms "processes" and "processors" are used interchangeably

Data exchange

timetime

P1

P1 P2 P3

P2

P3

Process 0

Process 1

Process 2

SerialSerial

MessageMessagePassingPassing

33

OpenMP

35

OpenMP Mission• The OpenMP Application Program Interface

(API) supports multi-platform shared-memory parallel programming in C/C++ and Fortran on all architectures, including Unix platforms and Windows NT platforms.

• Jointly defined by a group of major computer hardware and software vendors.

• OpenMP is a portable, scalable model that gives shared-memory parallel programmers a simple and flexible interface for developing parallel applications for platforms ranging from the desktop to the supercomputer.

35

36

OpenMP compiler choice

• gcc 4.40 or above– compile with -fopenmp

• Intel 10.1 or above– compile with –Qopenmp on Windows– compile with –openmp on linux

• PGI compiler– compile with –mp

• Absoft Pro Fortran– compile with -openmp

36

37

Sample openmp example

#include <omp.h>

#include <stdio.h>

int main() {

#pragma omp parallelprintf("Hello from thread %d, nthreads %d\n", omp_get_thread_num(), omp_get_num_threads());

}

37

38

serial-pi.c#include <stdio.h>

static long num_steps = 10000000;

double step;

int main ()

{ int i; double x, pi, sum = 0.0;

step = 1.0/(double) num_steps;

for (i=0;i< num_steps; i++){

x = (i+0.5)*step;

sum = sum + 4.0/(1.0+x*x);

}

pi = step * sum;

printf("Est Pi= %f\n",pi);

}

38

39

Openmp version of spmd-pi.c#include <omp.h>

#include <stdio.h>

static long num_steps = 10000000;

double step;

#define NUM_THREADS 8

int main ()

{ int i, nthreads; double pi, sum[NUM_THREADS];

step = 1.0/(double) num_steps;

omp_set_num_threads(NUM_THREADS);

#pragma omp parallel

{

int i, id,nthrds;

double x;

id = omp_get_thread_num();

nthrds = omp_get_num_threads();

if (id == 0) nthreads = nthrds;

for (i=id, sum[id]=0.0;i< num_steps; i=i+nthrds) {

x = (i+0.5)*step;

sum[id] += 4.0/(1.0+x*x);

}

}

for(i=0, pi=0.0;i<nthreads;i++)

pi += sum[i] * step;

printf("Est Pi= %f using %d threads \n",pi,nthreads);

}

39

Message Passing Interface (MPI)

41

MPI

• Is a library but not a language, for parallel programming

• An MPI implementation consists of– a subroutine library with all MPI functions– include files for the calling application program– some startup script (usually called mpirun, but not

standardized)

• Include the lib file mpi.h (or however called) into the source code

• Libraries available for all major imperative languages (C, C++, Fortran …)

41

42

General MPI Program Structure

MPI include file

#include <mpi.h>void main (int argc, char *argv[]){int np, rank, ierr;ierr = MPI_Init(&argc, &argv); MPI_Comm_rank(MPI_COMM_WORLD,&rank);MPI_Comm_size(MPI_COMM_WORLD,&np);/* Do Some Works */ierr = MPI_Finalize();}

variable declarations #include <mpi.h>void main (int argc, char *argv[]){int np, rank, ierr;ierr = MPI_Init(&argc, &argv); MPI_Comm_rank(MPI_COMM_WORLD,&rank);MPI_Comm_size(MPI_COMM_WORLD,&np);/* Do Some Works */ierr = MPI_Finalize();}

Initialize MPI environment

#include <mpi.h>void main (int argc, char *argv[]){int np, rank, ierr;ierr = MPI_Init(&argc, &argv); MPI_Comm_rank(MPI_COMM_WORLD,&rank);MPI_Comm_size(MPI_COMM_WORLD,&np);/* Do Some Works */ierr = MPI_Finalize();}

Do work and make message passing calls

#include <mpi.h>void main (int argc, char *argv[]){int np, rank, ierr;ierr = MPI_Init(&argc, &argv); MPI_Comm_rank(MPI_COMM_WORLD,&rank);MPI_Comm_size(MPI_COMM_WORLD,&np);/* Do Some Works */ierr = MPI_Finalize();}

Terminate MPI Environment

#include <mpi.h>void main (int argc, char *argv[]){int np, rank, ierr;ierr = MPI_Init(&argc, &argv); MPI_Comm_rank(MPI_COMM_WORLD,&rank);MPI_Comm_size(MPI_COMM_WORLD,&np);/* Do Some Works */ierr = MPI_Finalize();}

42

43

Sample Program: Hello World!

• In this modified version of the "Hello World" program, each processor prints its rank as well as the total number of processors in the communicator MPI_COMM_WORLD.

• Notes:– Makes use of the pre-defined

communicator MPI_COMM_WORLD.– Not testing for error status of routines!

43

44

Sample Program: Hello World!#include <stdio.h>#include “mpi.h” // MPI compiler header file

void main(int argc, char **argv) {

int nproc,myrank,ierr;

ierr=MPI_Init(&argc,&argv); // MPI initialization

// Get number of MPI processesMPI_Comm_size(MPI_COMM_WORLD,&nproc);

// Get process id for this processorMPI_Comm_rank(MPI_COMM_WORLD,&myrank);

printf (“Hello World!! I’m process %d of %d\n”,myrank,nproc);

ierr=MPI_Finalize(); // Terminate all MPI processes

}

44

45

Performance

• When we write a parallel program, it is important to identify the fraction of the program that can be parallelized and to maximize it.

• The goals are:– load balance– memory usage balance– minimize communication overhead– reduce sequential bottlenecks– scalability

45

46

Compiling & Running MPI Programs

• Using mvapich 1.1

1. Setting path, at the command prompt, type: export PATH=/u1/local/mvapich1/bin:$PATH

(uncomment this line in .bashrc)

2. Compile using mpicc, mpiCC, mpif77 or mpif90, e.g.mpicc –o cpi cpi.c

3. Prepare hostfile (e.g. machines) number of compute nodes:

Compute-0-0Compute-0-1Compute-0-2Compute-0-3

4. Run the program with a number of processor node:mpirun –np 4 –machinefile machines ./cpi

46

47

Compiling & Running MPI Programs

• Using mvapich 1.2

1. Prepare .mpd.conf and .mpd.passwd and saved in your home directory :

MPD_SECRETWORD=gde1234-3

(you may set your own secret word)

2. Setting environment for mvapich 1.2export MPD_BIN=/u1/local/mvapich2

export PATH=$MPD_BIN:$PATH

(uncomment this line in .bashrc)

3. Compile using mpicc, mpiCC, mpif77 or mpif90, e.g.mpicc –o cpi cpi.c

4. Prepare hostfile (e.g. machines) one hostname per line like previous section

47

48

Compiling & Running MPI Programs

5. Pmdboot with the hostfile mpdboot –n 4 –f machines

6. Run the program with a number of processor node:mpiexec –np 4 ./cpi

7. Remember to clean after running jobs by mpdallexitmpdallexit

48

49

Compiling & Running MPI Programs

• Using openmpi:1.21. Setting environment for openmpi

export LD-LIBRARY_PATH=/u1/local/openmpi/

lib:$LD-LIBRARY_PATH

export PATH=/u1/local/openmpi/bin:$PATH

(uncomment this line in .bashrc)

2. Compile using mpicc, mpiCC, mpif77 or mpif90, e.g.mpicc –o cpi cpi.c

3. Prepare hostfile (e.g. machines) one hostname per line like previous section

4. Run the program with a number of processor nodempirun –np 4 –machinefile machines ./cpi

49

50

Submit parallel jobs into torque batch queue Prepare a job script, say omp.pbs like the following

#!/bin/sh

### Job name

#PBS -N OMP-spmd

### Declare job non-rerunable

#PBS -r n

### Mail to user

##PBS -m ae

### Queue name (small, medium, long, verylong)

### Number of nodes

#PBS -l nodes=1:ppn=8

#PBS -l walltime=00:08:00

cd $PBS_O_WORKDIR

export OMP_NUM_THREADS=8

./omp-test

./serial-pi

./omp-spmd-pi

Submit it using qsubqsub omp.pbs

50

51

Another example of pbs scripts Prepare a job script, say scripts.sh like the following

#!/bin/sh

### Job name

#PBS -N Sorting

### Declare job non-rerunable

#PBS -r n

### Number of nodes

#PBS -l nodes=4

#PBS -l walltime=08:00:00

# This job's working directory

echo Working directory is $PBS_O_WORKDIR

cd $PBS_O_WORKDIR

echo Running on host `hostname`

echo Time is `date`

echo Directory is `pwd`

echo This jobs runs on the following processors:

echo `cat $PBS_NODEFILE`

# Define number of processors

NPROCS=`wc -l < $PBS_NODEFILE`

echo This job has allocated $NPROCS nodes

# Run the parallel MPI executable

/u1/local/mvapich1/bin/mpirun -v -machinefile $PBS_NODEFILE -np $NPROCS ./bubblesort >> bubble.out

Submit it using qsubqsub scripts.sh

51

Parallel Program Examples

53

Example 1: Estimation of Pi (OpenMP)

#include <omp.h>

#include <stdio.h>

static long num_steps = 10000000;

double step;

#define NUM_THREADS 8

int main ()

{ int i, nthreads; double pi, sum[NUM_THREADS];

step = 1.0/(double) num_steps;

omp_set_num_threads(NUM_THREADS);

#pragma omp parallel

{

int i, id,nthrds;

double x;

id = omp_get_thread_num();

nthrds = omp_get_num_threads();

if (id == 0) nthreads = nthrds;

for (i=id, sum[id]=0.0;i< num_steps; i=i+nthrds) {

x = (i+0.5)*step;

sum[id] += 4.0/(1.0+x*x);

}

}

for(i=0, pi=0.0;i<nthreads;i++)

pi += sum[i] * step;

printf("Est Pi= %f using %d threads \n",pi,nthreads);

}

53

54

Example 2a: Sorting – quick sort

• The quick sort is an in-place, divide-and-conquer, massively recursive sort.

• The efficiency of the algorithm is majorly impacted by which element is chosen as the pivot point.

• The worst-case efficiency of the quick sort, O(n2), occurs when the list is sorted and the left-most element is chosen.

• If the data to be sorted isn't random, randomly choosing a pivot point is recommended. As long as the pivot point is chosen randomly, the quick sort has an algorithmic complexity of O(n log n).

Pros: Extremely fast.Cons: Very complex algorithm, massively recursive

54

55

Quick Sort Performance

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1 2 4 8 16 32 64 128

Processes

Tim

e(se

c)

Processes Time

1 0.410000

2 0.300000

4 0.180000

8 0.180000

16 0.180000

32 0.220000

64 0.680000

128 1.300000

55

56

Example 2b: Sorting -Bubble Sort• The bubble sort is the oldest and simplest sort in use.

Unfortunately, it's also the slowest. • The bubble sort works by comparing each item in the

list with the item next to it, and swapping them if required.

• The algorithm repeats this process until it makes a pass all the way through the list without swapping any items (in other words, all items are in the correct order).

• This causes larger values to "bubble" to the end of the list while smaller values "sink" towards the beginning of the list.

56

57

Bubble Sort Performance

Processes Time

1 3242.327

2 806.346

4 276.4646

8 78.45156

16 21.031

32 4.8478

64 2.03676

128 1.240197

0

500

1000

1500

2000

2500

3000

3500

1 2 4 8 16 32 64 128

Processes

Tim

e (s

ec)

57

58

Monte Carlo Integration

• "Hit and miss" integration

• The integration scheme is to take a large number of random points and count the number that are within f(x) to get the area

58

59

Monte Carlo Integration

• Monte Carlo Integration to Estimate Pi

4

141

square inside hittingpt of no.

area shaded hittingpt of no.2

2

r

r

59

60

Example 1: omp

omp/test-omp.comp/serial-pi.comp/spmd-pi.cCompile program by the command: makeRun the program in parallel by./omp-spmd-piSubmit job to PBS byqsub omp.pbs

Example 3: Sorting

sorting/qsort.csorting/bubblesort.csorting/script.shsorting/qsort sorting/bubblesort

Submit job to PBS queuing system byqsub script.sh

Example 2: Prime

prime/prime.cprime/prime.f90prime/primeParallel.cprime/Makefileprime/machines

Compile by the command: makeRun the serial program by./primeC or ./primeFRun the parallel program bympirun –np 4 –machinefile machines ./primeMPI

Example 4: pmatlab

pmatlab/startup.mpmatlab/RUN.mpmatlab/sample-pi.m

Submit job to PBS byqsub Qpmatlab.pbs

60

Policy for using sciblade.sci.hkbu.edu.hk

62

Policy1. Every user shall apply for his/her own computer user

account to login to the master node of the PC cluster, sciblade.sci.hkbu.edu.hk.

2. The account must not be shared his/her account and password with the other users.

3. Every user must deliver jobs to the PC cluster from the master node via the PBS job queuing system. Automatically dispatching of job using scripts or robots are not allowed.

4. Users are not allowed to login to the compute nodes.

5. Foreground jobs on the PC cluster are restricted to program testing and the time duration should not exceed 1 minutes CPU time per job.

63

Policy (continue)6. Any background jobs run on the master node or

compute nodes are strictly prohibited and will be killed without prior notice.

7. The current restrictions of the job queuing system are as follows,– The maximum number of running jobs in the job queue is 8.– The maximum total number of CPU cores used in one time

cannot exceed 512.

8. The restrictions in item 7 will be reviewed timely for the growing number of users and the computation need.