This is an example of a server that consumes data from clients concurrently and logs the data in a high-performance fashion. The application can handle up to 5 concurrent clients on port 3000. The client(s) can write any number of valid data lines (9 digit number with leading 0's if necessary + a newline character) to the server and optionally send "terminate" + newline to shut down the server. Invalid data terminates the connection/thread that encountered it. The numbers are logged in the file numbers.log and duplicate numbers are discarded. Every 10 seconds a report is printed to stdout showing deltas for the count of unique and duplicate numbers, in addition to the total number of unique numbers.
The average observed throughput is about 500K numbers per second, but the bottleneck could live in the client. Memory usage should sit steady at around 1GB, mainly because of duplicate tracking.
Testing machine: 8GB RAM, 2.3GHz Intel i5, 256GB SSD.
Open the project folder in IntelliJ. Assuming IntelliJ is already configured with your JDK, running the server should be as simple as using the green arrow in the Main class or using the Run dropdown. Building is done through the Build dropdown.
Navigate to the project folder and run ./gradlew build to build the project. Run
./gradlew shadowJar to create a complete runnable jar. Run
java -jar build/libs/sanchez-server.jar to run the server.
To run a few (poorly implemented) tests through IntelliJ, they are located in the MainTest class in the test directory. The most interesting one is probably testThroughput(). Or you can copy one of the client files in the test directory, and run it through a basic main method somewhere else (to test outside of IntelliJ). Or you can make your own client.
- Clients are on the local area network
- Clients connect through port 3000 and use TCP/IP
- Server is run on Linux (OSX should work too though)
- Java JDK 11 is installed
- Machine has a couple GB of hard disk space to hold numbers.log