Management and Analysis of Physics Datasets (Mod. B) - 2020/2021

First step

Clone this repo if has not already been done: git clone https://github.com/sirCamp/ManagementAndAnalysisOfPhysicsDatasetsB.git All the lectures will be pushed on the repo. Update it before each lecture.

Docker Users

It is possible to run all the exercises using docker. This is particularly usefull if you are experiencing problems with virtual or you don't want to go to the trouble of setting up the environment.

In this way, the only thing you need to install is docker itself (instructions for Mac, Windows and linux can be found here)

Single container

Run the following command, from inside ManagementAndAnalysisOfPhysicsDatasetsB directory, to create a container with dask, jupyter and others packages already installed. The first time you run it may take a while because it needs to download all the necessary components and libraries.

Due to the large amount of data that have to be download, please do it before the lectures

However, the next times you will need this container it will be created instantly since all the components are cached.

docker run --rm -it \
        -p 8888:8888 \
        -p 8787:8787 \
        -v $PWD:/opt/workspace \
        sircamp/mapd:parallel

Once it has been created, you will see some logs from the jupyter server. At this point you can connect to jupyter from your browser typing localhost:8888. Make sure that other jupyter are not using this port.

Docker cluster

We can simulate multiple nodes by using multiple docker containers:

A container hosting jupyter server and acting as the client
A container hosting the dask-scheduler (aka the master node)
A/multiple container/s hosting the dask-worker nodes that will connect to the scheduler.

As for the Spark part, this setup is simulating what you experience on a real cluster.

This cluster can be spawned with

docker compose up --scale dask-worker-custom=0 --scale dask-worker-gpu-big=0

By default only one worker is created. If you want to use N workers you can scale the cluster with

docker compose up --scale dask-worker=N

Please pay attention when you want to scale up the workers number, each worker is designed to have 1 core - 1 thread -1Gb of RAM. Do not exceed your machine resources

Please note that for some of you that use older version of docker have to install the docker compose (here you can find how), and the command should be replaced by docker-compose

If you want to add "on the fly" new nodes with particular resource to the cluster (on a different CMD):

sudo docker-compose run  -e DASK_WORKER_RESOURCES=3 -e DASK_WORKER_RESOURCES_NAME=GPU dask-worker-custom

sudo docker-compose run  dask-worker-gpu-big

Virtual Machine Users

If you prefer to use you VM in VirtualBox or your own machine before the lectures please install the following packages:

The graphviz package via apt-get: sudo apt-get install graphviz.
The following python packages:

pip3 install  pandas \
   numpy \
   "dask[complete]" \
   "dask[array]"    \
   "dask[dataframe]"  \
   "dask[diagnostics]"  \
   "dask[distributed]"  \
   dask-ml \
   scikit-learn \
   numba \
   notebook \
   jupyterlab \
   matplotlib \
   Pillow \
   seaborn \
   graphviz \
   xgboost \
   joblib \
   pyarrow \
   tensorflow==2.4.1 \
   keras \
   torch \
   scikeras==0.3.3 \
   jupyter-server-proxy \

Check if all is running by typing dask-scheduler --help in your command line.

Lectures

Lecture 1

Introduction on what is Dask
Parallel computing base
~~Distributed computing base~~
Examples
Exercises

Lecture 2

Distributed computing base
Dask Examples
Distributed ML
Exercises discussion

Lecture 3

Lecture 2 remaining
Dask Bag
Dask dataframe
Map Reduce Concepts
Exercises

SaverioMonaco / ManagementAndAnalysisOfPhysicsDatasetsB