XClose

HPL - High Performance Linpack Benchmark:

Testing on multiple computing platforms with Docker

Using Ubuntu, I have created a Docker image that installs the dependencies, then downloads, compiles, configures and runs the HPL benchmark. The benchmark measures performance in solving a dense system of linear equations; the user is able to scale the size of the problem.

The computing platforms I will test on are as follows:

  1. Mac laptop
  2. Azure VM 1 - Standard D2s v3 (2 vcpus, 8 GB memory)
  3. Azure VM 2 - Standard NC6 (6 vcpus, 56 GB memory)

Note: I created a custom Make file to work with the dependencies installed via apt and apt-get

%%writefile Dockerfile
FROM ubuntu:18.04

# Install dependencies  
RUN apt-get update
RUN apt-get install build-essential emacs -y
RUN apt install libopenmpi-dev -y
RUN apt install openssh-server -y
RUN apt-get install libatlas-base-dev gfortran -y

# Download HPL
RUN wget http://www.netlib.org/benchmark/hpl/hpl-2.3.tar.gz
RUN tar xf hpl-2.3.tar.gz
RUN mv hpl-2.3 hpl

# Compile the benchmark with custom Make file
COPY Make.ubuntu /hpl/Make.ubuntu
WORKDIR "/hpl"
RUN make arch=ubuntu
WORKDIR "/hpl/bin/ubuntu"

# Custom config file
COPY HPL.dat HPL.dat

# Run the benchmark
CMD mpirun -np 4 --allow-run-as-root ./xhpl

Overwriting Dockerfile

Setup the HPL.dat configuration file for benchmarking

Each line of this config file is explained in here.

Here I keep many of the default settings, but change the following:

  1. N - set the number of linear equations being solved to 10,000
  2. P and Q - since I am using 4 MPI ranks when running the benchmarks (see the CMD in the Dockerfile), I set a 2 x 2 process grid
%%writefile HPL.dat
HPLinpack benchmark input file
Innovative Computing Laboratory, University of Tennessee
HPL.out      output file name (if any)
6            device out (6=stdout,7=stderr,file)
1            # of problems sizes (N)
10000         Ns
1            # of NBs
232          NBs
0            PMAP process mapping (0=Row-,1=Column-major)
1            # of process grids (P x Q)
2            Ps
2           Qs
16.0         threshold
1            # of panel fact
2            PFACTs (0=left, 1=Crout, 2=Right)
1            # of recursive stopping criterium
4            NBMINs (>= 1)
1            # of panels in recursion
2            NDIVs
1            # of recursive panel fact.
1            RFACTs (0=left, 1=Crout, 2=Right)
1            # of broadcast
1            BCASTs (0=1rg,1=1rM,2=2rg,3=2rM,4=Lng,5=LnM)
1            # of lookahead depth
1            DEPTHs (>=0)
2            SWAP (0=bin-exch,1=long,2=mix)
64           swapping threshold
0            L1 in (0=transposed,1=no-transposed) form
0            U  in (0=transposed,1=no-transposed) form
1            Equilibration (0=no,1=yes)
8            memory alignment in double (> 0)

Overwriting HPL.dat

%%bash
docker build -t edwardchalstrey/hpl_benchmark .

Step 1/15 : FROM ubuntu:18.04
 ---> 1d9c17228a9e
Step 2/15 : RUN apt-get update
 ---> Using cache
 ---> d9ff41fcf2c7
Step 3/15 : RUN apt-get install build-essential emacs -y
 ---> Using cache
 ---> 1ae04d308aab
Step 4/15 : RUN apt install libopenmpi-dev -y
 ---> Using cache
 ---> 7b2bce57546a
Step 5/15 : RUN apt install openssh-server -y
 ---> Using cache
 ---> 66f0dcc1ec2f
Step 6/15 : RUN apt-get install libatlas-base-dev gfortran -y
 ---> Using cache
 ---> 0c32d16a01c2
Step 7/15 : RUN wget http://www.netlib.org/benchmark/hpl/hpl-2.3.tar.gz
 ---> Using cache
 ---> db1c09a4bd4e
Step 8/15 : RUN tar xf hpl-2.3.tar.gz
 ---> Using cache
 ---> 723e4047ab5c
Step 9/15 : RUN mv hpl-2.3 hpl
 ---> Using cache
 ---> bd3415912c99
Step 10/15 : COPY Make.ubuntu /hpl/Make.ubuntu
 ---> Using cache
 ---> 3308c52d2b49
Step 11/15 : WORKDIR "/hpl"
 ---> Using cache
 ---> ebbfc230b7fc
Step 12/15 : RUN make arch=ubuntu
 ---> Using cache
 ---> 9fd80ebb632b
Step 13/15 : WORKDIR "/hpl/bin/ubuntu"
 ---> Using cache
 ---> 729147c5ad37
Step 14/15 : COPY HPL.dat HPL.dat
 ---> Using cache
 ---> df0731f40190
Step 15/15 : CMD mpirun -np 4 --allow-run-as-root ./xhpl
 ---> Using cache
 ---> 0b10062c35f9
Successfully built 0b10062c35f9
Successfully tagged edwardchalstrey/hpl_benchmark:latest

%%bash
docker push edwardchalstrey/hpl_benchmark

The push refers to repository [docker.io/edwardchalstrey/hpl_benchmark]
6f3f7e77eef4: Preparing
77d796561be4: Preparing
d6a2438a68fe: Preparing
76db63a1ea23: Preparing
031701f75160: Preparing
dcfb774847e1: Preparing
91d0f71d8391: Preparing
04d308d03718: Preparing
204467103833: Preparing
ab748e27090c: Preparing
7a4226e9f5db: Preparing
2c77720cf318: Preparing
1f6b6c7dc482: Preparing
c8dbbe73b68c: Preparing
2fb7bfc6145d: Preparing
91d0f71d8391: Waiting
04d308d03718: Waiting
204467103833: Waiting
ab748e27090c: Waiting
7a4226e9f5db: Waiting
2c77720cf318: Waiting
1f6b6c7dc482: Waiting
c8dbbe73b68c: Waiting
2fb7bfc6145d: Waiting
dcfb774847e1: Waiting
031701f75160: Layer already exists
77d796561be4: Layer already exists
76db63a1ea23: Layer already exists
6f3f7e77eef4: Layer already exists
d6a2438a68fe: Layer already exists
dcfb774847e1: Layer already exists
91d0f71d8391: Layer already exists
204467103833: Layer already exists
04d308d03718: Layer already exists
ab748e27090c: Layer already exists
7a4226e9f5db: Layer already exists
1f6b6c7dc482: Layer already exists
2c77720cf318: Layer already exists
c8dbbe73b68c: Layer already exists
2fb7bfc6145d: Layer already exists
latest: digest: sha256:70b5800b688b26ccf6e9de67e3d37ea8659c0e1ca839d0266fead60b750e315e size: 3465

Benchmark on different computing platforms

  1. Mac laptop
  2. Azure VM 1 - Standard D2s v3 (2 vcpus, 8 GB memory)
  3. Azure VM 2 - Standard NC6 (6 vcpus, 56 GB memory)

docker run edwardchalstrey/hpl_benchmark

from IPython.display import HTML, display
import tabulate

headers = ["Platform", "Time (s)", "Gflops"]
results_1 = ["Mac laptop", 16.87, 3.9520e+01]
results_2 = ["Azure VM 1", 80.36, 8.2983e+00]
results_3 = ["Azure VM 2", 17.16, 3.8856e+01]
display(HTML(tabulate.tabulate([headers, results_1, results_2, results_3], tablefmt='html')))
Platform Time (s)Gflops
Mac laptop16.87 39.52
Azure VM 180.36 8.2983
Azure VM 217.16 38.856