Code and repo for the paper titled "SpotWeb: Running Latency-sensitive Distributed Web Services on Transient Cloud Servers" from ACM HPDC 2019. See paper at http://lass.cs.umass.edu/publications/pdf/hpdc19.pdf
For SpotWeb to run, the following python libraries are needed:
a- boto3 ( for connecting with AWS)
b-flask
c-rpy2 (for the splines predictor)
d- pandas
e-flask_restful
f-pickle (for storing the output from the experiments)
g- cvxportfolio
h- cvxpy
For plotting:
a- seaborn
b-matplotlib
There are six folders in the repo:
a- The SpotWeb folder contains all the code used to run the core of SpotWeb, the MPO optimizer, and the simulations. SpotWeb is now released as a library. For example usage, please check the simulation code or contact us. To connect SpotWeb to a cloud provider, you will need to design an interface to connect to the Cloud provider.
c- The Autoscaler folder contains the code for an autoscaler that was used for the paper. In addition, we add a pointer to a few more autoscaler implementations.
d- The data folder contains most of the data we collected from monitoring AWS spot markets.
e-The loadbalancer folder contains the code used for managing the load balancing in SpotWeb including the automation and the scripts required to change the loadbalancing coefficients.
The German Wikipedia images are available here (http://zenky.cs.umu.se/PEAS/)