This repository contains a fog application that simulates an adaptive speed limit use-case to demonstrate reliable message delivery. It consists of a zmq-client.py script to simulate speed-limit sensors and a zmq-server.py script to simulate a processing component in the cloud. The ZeroMQ messaging library is used to facilitate communication between components.

The domain of this Proof of concept is an Autobahn sensor network, in which roadway sensors are reading the speed of passing cars, sending these readings to a centralized system. The centralized system will then aggregate these speed readings into an average speed within a 5 second window and respond to each client with the most recent average speeds. The idea being that this information can be used for setting adapative speed limits.

The client/server architecture combines multple strategies to ensure reliable message delivers

All communications originating from the client or server must be acknowledged by the recieving party.

• The clients are responsible for generating Speed messages, which conatin the clients id, and a sequence number. The Server will respond with an acknowledgement to every message. The Client is responsible for buffering and resending unacknowledged messages.

• The servers generate Average speed messages to send to each client. The server retains a hasmap for each connected client and its last acknowledged sequence number and buffers all messages greater than the minimum client sequence.

These mechanisms ensure reliable deliver of messages in continuous connectivity scenarios, the strategies below will describe how reliability is ensured in disconnection scenarios.

The Server architecture is based off the Binary Star example provided by zmq documentation. This server topology is effectively a Primary/Replica server setup, in which the client is designed to failover to the replica if a request timeout occurs on the primary server.

On top of the Binary Star implementation we extended the functionality so that primary server is responsible for sending updates to the replica server so that in the case of a failover to the replica, unsent messages and sequence/acknowledgement states will be up to date providing transparent failover from the clients point of view.

The replication is done by the primary writing its state to a file and rsyncing this information to the replicas working directory. This approach is inspired by the Write Ahead Log style backup that Postgres uses, or the Append only Log which Redis uses in which a Hard drive based backup can be replayed to recover the systems state in case of a complete downtime.

From the client side all unacknowledge messages are buffered in memory until sent and properly acknowledged from the server side.

With the combination of Acknowledgements, Primary/Replica avaliability and replication, and Message buffering the entire system is robust enough to effectively continuously serve its intended function with 2/4 systems enitrely crashed or disonnected.

Start the primary and backup cloud servers with the following commands:

python3 zmq-server.py --help
usage: zmq-server.py [-h] [-p | -b] [-ip IP] [-u USERNAME] [-rd RSYNCDIR]

optional arguments:
  -h, --help            show this help message and exit
  -p, --primary         # Server Role
  -b, --backup          # Server Role
  -ip IP, --ip IP       # Ip of Peer, default localhost
  -u USERNAME, --username USERNAME     # SSH Username, default petbangert
  -rd RSYNCDIR, --rsyncdir RSYNCDIR    # Target Dir for Rsync, default ~/fc-ss2021/

Start sensors with the client script by passing a unique id:

python3 zmq-client.py --help
usage: zmq-client.py [-h] [-id ID] [-ip1 IP1] [-ip2 IP2] [-u USERNAME]

optional arguments:
  -h, --help            show this help message and exit
  -id ID, --id ID       # Client ID
  -ip1 IP1, --ip1 IP1   # IP of Server 1, default localhost
  -ip2 IP2, --ip2 IP2   # IP of Server 2, default localhost
  -u USERNAME, --username USERNAME  # SSH Username, default petbangert

Terminal 1

$ python3 zmq-server.py --primary --username USERNAME

Terminal 2

$ python3 zmq-server.py --backup --username USERNAME

Terminal 3

$ python3 zmq-client.py --id 3

A video demo can be seen in the media/demo.mp4 file.