credibility_allocation's Introduction

Credibility Allocation

Simulate credibility allocation using the Google Cluster Data.

In this repo you will find the code to aggregate users requests, discretize their needs, simulate the credibility allocation and run some experiments on it.

Usage

Download the Google dataset following the instructions here.

Clone this repo:

git clone [email protected]:hugombarreto/credibility_allocation.git

Install the dependencies:

pip install -r credibility_allocation/requirements.txt

Run it for usage information:

python credibility_allocation --help

For further help in a particular subcommand you can run:

python credibility_allocation SUBCOMMAND --help

Subcommands

Here we detail the purpose of each subcommand. In an usual workflow they are likely used in the order they appear here (except for the multicore_simulate_allocation which is in fact an alternative to the simulate_allocation).

aggregate_user_events:	Inspect dataset events aggregating cpu and memory requests for each user in a separate file for each.
generate_user_needs:	Aggregates all resource requests so that we have, at a given timestamp, the amount of resources needed by the user.
sample_needs:	Sample needs for each user following a period specified. Creates a single file with all users types for every timestamp.
simulate_allocation:	Simulate allocation using sampled types created with the `sample_needs`. It generates a file with the allocations for every user on every iteration.
multicore_simulate_allocation:	Alternative to the `simulate_allocation` using multiple CPU cores. It monitors the memory usage so as to avoid swapping.
experiments:	Runs a set of experiments using the simulation output.

Recommend Projects