Aegean

The source code and scripts for the implementation and evaluation of the Aegean replication framework.

Implementation Source Code

Aegean source code is under src directory. Two important folder under this directory are BFT and Applications.

• BFT folder mainly contains replication infrastructure. Here is explanation of some subfolders under BFT:

subfolder name	description
exec	implements executor which executes request in sequential and parallel mode
filter	implements the module which authenticates incoming client requests
order	module which orders the client requests before the execution of requests in sequential mode
verifier	implements verifier process which decides the resulting state after execution in parallel mode

• Applications directory mainly contains experimental applications which runs on top of replication infrastructure. For example, Benchmark folder contains implementation of microbenchmarks and folder prefixed with tpcw contains implementation of TPC-W benchmark such as tpcw_webserver (web servers in TPC-W benchmark).

Evaluation Scripts

Experiments are run with the scripts under experiments directory. Here are some scripts with their functions:

script name	description
start_client.py	starts requested number of clients which will send requests to middle service for microbenchmarks
start_sequential_middle.py	starts executor, order and filter nodes to start a replicated middle service in sequential mode for microbenchmarks
start_sequential_backend.py	starts executor, order and filter nodes to start a replicated backend service in sequential mode for microbenchmarks
start_middle.py	starts executor, verifier and filter nodes to start a replicated middle service in parallel mode for microbenchmarks
start_backend.py	starts executor, verifier and filter nodes to start a replicated backend service in parallel mode for microbenchmarks
start_tpcw_client.py	starts requested number of clients which will send requests to tpc-w web server for TPC-W benchmark
start_sequential_tpcw_server.py	starts a tpc-w web server in sequential mode which includes starting executor, order and filter nodes
start_sequential_tpcw_db.py	starts a tpc-w database server in sequential mode which includes starting executor, order and filter nodes
start_tpcw_server.py	starts a tpc-w web server in parallel mode which includes starting executor, verifier and filter nodes
start_tpcw_db.py	starts a tpc-w database server in parallel mode which includes starting executor, order and filter node

Other high-level scripts which calls the scripts described above are used to initiate experiments. These high-level scripts also checks for completion and calculate the throughput. Here are some high level scripts with their usage and explanations:

script and example usage	effect of example usage
./pipelinedSequential_super.py --mode ps 1 4 seqMicro	runs microbenchmarks in pipelined sequential mode and calculates the performance. Number of clients for this experiments starts from 1 and goes until there is no throughput increase or number of clients is greater than 4
./super_script_tpcw.py --mode p 128 1024 tpcwTest	runs tpc-w benchmark in parallel mode and calculates the performance. Number of clients for this experiments starts from 128 and goes until there is no throughput increase or number of clients is greater than 1024

Note: Size of the requests(small, medium,large) for microbenchmarks in start_client.py with REQUEST_SIZE variable (100, 1000, 10000).

Other than scripts, this directory also contains configuration files for the nodes will be started. Following are the list of files and their explanations:

config file	description
clients	IP or alias of machines where client process will be run on
filters	IP or alias of machines where middle service filter nodes will be run on
orders	IP or alias of machines where middle service order nodes will be run on for sequential mode
verifiers	IP or alias of machines where middle service verifier nodes will be run on for parallel mode
execs	IP or alias of machines where middle service executor nodes will be run
clients.backend	IP or alias of clients of backend (so should be same IPs with execs file)
filters.backend	IP or alias of machines where backend service filter nodes will be run on
orders.backend	IP or alias of machines where backend service order nodes will be run on for sequential mode
verifiers.backend	IP or alias of machines where backend service verifier nodes will be run on for parallel mode
execs.backend	IP or alias of machines where backend service executor nodes will be run on

Note: The configuration defined in these files should be converted to properties files such as test.properties and test.properties.backend which will be consumed by the processes while starting. The script calledconfig.sh can be used to ease this conversion. This script asks some other input regarding mode and number of tolerated failures etc. and reads the config files explained above to create properties files which define the setting for Java processes.