# Cloudbench

This is a project which tries to benchmark and compare multiple public
cloud providers. It also tries to identify the bottleneck resources in
the cloud platform for applications.


# Setup
Put your keys in the ./config/ folder.  For Azure you would need to
create two keys:

* cloud.pem
* cloud.key

# Running the experiment

Go to the script folder and use the run.sh script.  The default help is pretty
self explanatory:

>   Please add the configs that you want to run to the printConfigs
>   function inside the script.  The syntax is:
>
>   > configFor "Experiment" "InstanceType" "Machine Count" "Disk Type"
>
>   1) Experiment: any one of: tpcds, tpch, tera, spark
>
>   2) Instance type: any of the instance types in Amazon
>
>   3) Instance count: number of instances in the cluster
>
>   4) Disk type: ebs or empty string
>
>   By default the disks are set to be 2x250GB of gp2 type per instance.  
>   Feel free to change that within this file.

<!--
cloud.key should be a 2048bit RSA key.  You can generate Azure keys with
the openssl command or just the supplied makefile:

> make azure_keys

Also make sure that the permission of *cloud.key* is set to 600.

# Cloud specific notes

## Azure
* Because of plethora of random objects that Azure creates for you, as of
now, it is not possible to "cleanly" delete a topology.  This is a work
in progress and any feedbacks are welcome


## Examples

To run a specific benchmark you can use the 'bin/cb' binary.  For
example:

> ./cb --benchmark=ipref --setup --teardown


This command would first setup the environment specified in config.xml
for running the iperf benchmark located in cloudbench/benchmarks/iperf.
Then it would run the main.py script for benchmarking, and afterwards it
would teardown the environment.  If the environment is to be persisted
for next runs, you can avoid passing --teardown to cb.

## Benchmark format

All the benchmarks are located in the cloudbench/benchmarks/ folder.
To create a new benchmark, e.g., stress_test , you would need to create
a new folder called stress_test in the benchmarks folder.  At least two
files are required:

* *config.xml* which specifies the environment configuration, e.g.,
  virtual machines, virtual networks, etc.
* *main.py* where the benchmarking script is run in the context of the
  environment.

For an example, have a look at cloudbench/benchmarks/iperf.
-->