Remote auctions application

Example application demonstrating key ideas for "Economy Of Things" (EoT) use cases. This example demonstrates both the remote deployment of auction and settlement services that can be discovered by trading actors. The reote application consists of three main components.

• TradingSystem listens on port 2552 and starts potentially several AuctionParticipantActor instances that generate SingleUnitAskOrder and SingleUnitBidOrder instances and send them to the auction service for processing. In practice, I would anticipate some IoT-enabled device to have a JVM process with a TradingSystem with a single AuctionParticipantActor instance. TODO: use dccker and docker-compose to demonstrate how the system would work with N separate TradingSystem instances each running inside its own JVM process in a separate container.
• AuctionSystem listens on port 2553 and starts an AuctionActor instance that provides services for matching individual buyers and sellers at specific prices and quantities. The auction service generates streams of SpotContract instances which are sent to the settlement service for further processing.
• SettlementSystem listens on port 2554 and logs the received streams of SpotContract instances. At some point the SettlementSystem might interact with a Blockchain as part of the settlement service.

Each of the actor systems has its own configuration file located in the resources directory. The TradingSystem uses the trading.conf; AuctionSystem uses the auction.conf; and SettlementSystem uses the settlement.conf. All three configuration files share basic settings via common.conf. The common.conf enables remoting by installing the RemoteActorRefProvider and chooses the default remote transport. Note that when deploying the services on multiple machines you will need to change the default IP address with the real IP addresses.

The SettlementActor does not really illustrate anything exciting: it simple logs out each of the Fill instances it receives. The RemoteAuctionServiceActor takes a String path as constructor parameter. This is the full path, including the remote address of the settlement service.

"akka.tcp://[email protected]:2554/user/settlement"

Observe how the actor system name of the path matches the remote settlement system’s name, as do IP address and port number. As always, top-level actors are always created below the "/user" guardian, which supervises them.

Upon creation the RemoteAuctionServiceActor instance sends an Identify message to the actor selection of the path. The remote LoggingSettlementActor actor will reply with an ActorIdentity message containing its ActorRef. Note that Identify is a built-in message that all Akka Actor instances understand, and automatically reply to, with a ActorIdentity. If the RemoteAuctionServiceActor is unable to identify the remote settlement service, it will retry after some scheduled timeout duration.

Once the RemoteAuctionServiceActor has the ActorRef of the remote settlement service it can monitor it. The remote settlement system might be shutdown and later started up again, in which case the RemoteAuctionServiceActor would receive the Terminated message which prompts it to retry the identification process in orde to establish a connection to the new remote settlement system.

Running the remote applications

Let's run each of the actor systems remotely in separate JVM processes. Start the settlement system by opening up a new terminal window and running the following.

sbt "run-main RemoteSettlementServiceApp"

This should generate some generic logging cruft followed by a message indicating that the settlement service has been activated. The settlement service is the core service on top of which everything else depends. If the service is down or otherwise unreachable then the auction service can not operate and the participants cannot settle transactions.

Once the settlement system has started, start the auction system by opening up a new terminal window and running the following.

sbt "run-main RemoteAuctionServiceApp"

Again, this should generate some logging cruft followed by a message indicating that the auction service is operational. Finally, start the trading system by opening up a new terminal window and running thw following.

sbt "run-main org.economicsl.auctions.remote.RemoteAuctionExampleApp Trading"

Once the trading system is operational, you should be able to see SpotContract instances being logged to the terminal window handling the remote settlement service.