LHCb grid simulation

Goals and problems

Developing smarter algorithms for

1. Data management
2. Predicting anomalies
3. Job scheduling/matchning

This research focuses on real interaction of

1. CPU resources
2. Networks
3. Storages

Job workflow

Installation process

In this project we use two additional libraries. SimGrid framework is used for the purposes of simulation flow and YAML cpp is used for parsing input files containing information about jobs and datasets. After installation of SimGrid and YAML-cpp, clone this project:

git clone https://github.com/skygrid/grid_simulation.git

and run in the command line:

mkdir build
cd build
cmake .. 
make
./CSim2Sim --cfg=tracing:yes --cfg=tracing/platform:yes --cfg=tracing/filename:<path to trace file> --cfg=maxmin/concurrency_limit:100000 --cfg=storage/max_file_descriptors:220000"

Change config.yml to customize paths to the following files:

platform: platform/platform.xml           ## Contains grid topology
deployment: platform/test-deployment.xml  ## Defines `behaviour` of sites 
out.txt: out.txt                          ## Jobs metrics are written here 
jobs: InputData/little_jobs.yml           ## Queue of waiting jobs
input: InputData/little_data.yml          ## Datasets stored in all tiers 

Docker

If you feel free with Docker you can play around with grid_simulation in container.

You just need to pull images DockerHub repository:

docker pull kenenbek/grid_simulation:february

Or you can build grid_simulation image based on Dockerfile on your own:

docker build -t <container name> .

This image contains all necessary software to run a simulation. For this, clone this repository. This creates /grid_simulation directory with scripts and input data of the simulation:

git clone https://github.com/skygrid/grid_simulation.git

The docker image has a CMD instruction which starts an execution of run.sh file. This file is in your /grid_simulation directory. To run the simuation use the command:

docker run -v $(pwd)/grid_simulation:/grid_simulation <container name>

Results of the simulation are in your /grid_simulation directory.

Cern OpenStack Instructions Using Docker

1. Create instance on openstack.cern.ch. Use one of the Ubuntu 14+ images. These images are available on cloud-images.ubuntu.com.

2. Log in your instance and run the command:

   sudo apt-get update

3. Install Docker using the instructions. If your are non-root user, run the following commands:

   sudo groupadd docker
   sudo usermod -aG docker $USER
   sudo service docker restart

   Verify the installation by running:

   sudo docker run hello-world
   sudo docker ps -a

   These commands should work without errors.

Output

In the wake of simulation you should look for out.txt and trace.csv (located in simgridoutput directory) files. The first one contains all relevant information about job execution process:

+-----------+------+---------------+--------------+-----------+------------+-----------+--------------+
| JobId     | Done | Submission(s) | Schedule (s) | Start (s) | Finish (s) | Tier name | JobType      |
+-----------+------+---------------+--------------+-----------+------------+-----------+--------------+
| 141647523 | 1    | 0             | 10.5645      | 56.2571   | 14974.7336 | CERN-PROD | MCSimulation |
+-----------+------+---------------+--------------+-----------+------------+-----------+--------------+

The second one does information about resource(CPU-cores, network, etc.) usage:

+---------------------+----------+-----------+-----------+----------+--------------+
| Resource name       | Variable | Start     | End       | Duration | Value        |
+---------------------+----------+-----------+-----------+----------+--------------+
| CERN-RAL-LHCOPN-001 | traffic  | 17483.121 | 17535.824 | 52.702   | 342164520177 |
+---------------------+----------+-----------+-----------+----------+--------------+

Also

You could check presentation about LHCb simulation and main current results in jupyter notebook.