Installing SCIP & PySCIPOpt on an EC2 instance

What is SCIP, and PySCIPOpt?

SCIP is a free solver for Mixed Integer Programming (MIPS), and Mixed Integer NonLinear Programming (MINLP). It is written in C, and callable by C++. PySCIPOpt is a python wrapper around these same functions. While scipy has some nice built in optimization techniques, the inputs take in vectorized floating point integers. If you want to model a binary outcome, there are some hacks you can implement, but you’re largely out of luck.

Why is installing them a problem?

Installation of scientific computation libraries is notoriously difficult. I found myself mightily struggling with installing SCIP on an EC2 instance. I found that I was forced to use SCIP, because of its support for mixed integer programming (MIP) problems. There are other commercial options, but SCIP is free! The whole process took about a day to work through. I can only share my experience tyring to set it up on an EC2 instance.

There are some weird extenuating circumstances at play here that make the whole experience a bit more painful than it ought to be. SCIP requires the a relatively recent version of cmake, and requires boost. Both of these libraries, for whatever reason, do not play very nicely with the EC2 OS.

In any case, here’s what I did. Start in a root directory, and install as many of your system depdency libraries as you can using basic package managers:

# Install base dependencies
sudo pip3 install Cython
sudo apt-get install -y build-essential curl file git linuxbrew-wrapper libblkid-dev e2fslibs-dev libboost-all-dev libaudit-dev libspatialindex-dev
brew install icu4c
brew upgrade icu4c
brew install boost
brew upgrade boost

The next step is to install cmake. Simply installing cmake from the command line will install an older version of cmake. Thus, we need to clone and manually install it ourself.

# Install cmake
wget https://github.com/Kitware/CMake/releases/download/v3.17.2/cmake-3.17.2.tar.gz 
tar -xvf cmake-3.17.2.tar.gz 
cd cmake-3.17.2
./bootstrap
make
sudo make install
cd ..
rm cmake-3.17.2.tar.gz 
rm -rf cmake-3.17.2

Next, we install the SCIP optimization suite. The optimization suite really simplifies the whole installation process. I assume that the tar file is already installed. To my knowledge, there is no way to just wget the installation suite. The link can be found here.

# Download & Install SCIP using scipoptsuite
grep scipoptsuite-7.0.0.tgz -- check that copy of the installation suite is here.
tar -xvf scipoptsuite-7.0.0.tgz
rm scipoptsuite-7.0.0.tgz

cd scipoptsuite-7.0.0
mkdir build
cd build/
cmake ..
cd scip/
make
make check
sudo make install

Finally, install PySCIPOpt,

# Install PySCIPOpt
cd
git clone https://github.com/SCIP-Interfaces/PySCIPOpt.git
cd PySCIPOpt/
SCIPOPTDIR=./ python setup.py install

In the event you still have issues running PySCIPOpt, and see an error regarding a missing libscip.so file. You should copy-paste your libscip.so file installed by the SCIP optimization suite, likeso.

sudo cp /usr/local/lib/libscip.so.7.0 /lib/x86_64-linux-gnu/ # copy-paste libscip.so.7.0 file

Hopefully, this helps someone else as much as I know it would have helped me!