StreamLet Project

Group Number 3

Name	Number
Diogo Sousa	59792
Denis Bahnari	59878
Bruno Faustino	59784

---

Project Description

The project was developed in Java, more specifically in version 25. The entire code structure is in the folder StreamLet/, which corresponds to the project root.

---

Compilation Instructions

Inside the StreamLet/ folder, there is a Makefile that automates the build process. To compile the project, just run on the terminal:

make

---

Running the System

Inside the StreamLet/out/ folder, there is the main Streamlet class, which can be started with the following command:

java -cp out Streamlet <id>

where id represents the node identifier. The project currently supports exactly 5 nodes. In order for the system to work properly, it is necessary to start up to five different instances, each with a distinct ID (for example, from 0 to 4):

java -cp out Streamlet 0
java -cp out Streamlet 1
java -cp out Streamlet 2
java -cp out Streamlet 3
java -cp out Streamlet 4

Another easy way to start the system is by running the provided shell script(Linux Ubuntu 24):

./run_all.sh

This script automatically launches all five nodes (IDs 0–4) in separate processes, allowing the project to start and operate correctly without manual intervention.

If permission is denied, execute:

chmod +x run_all.sh

To execute the client process run:

java -cp out StreamletClient

Features

All required features were implemented along with several additional enhancements:

1. Persistence: The blockchain maintains disk-based persistence for each node in the output folder using a dual-file approach. The primary file contains a snapshot of the in-memory data structures, updated every N epochs (currently 10). A secondary log file records all operations between snapshots. This design ensures full crash recovery capability.

2. Client Software: A dedicated client application (StreamletClient class) enables external clients to submit transactions to any active server node. The node incorporates these transactions into blocks, replacing the default dummy transaction mechanism.

3. Synchronization and Recovery: A catch-up mechanism allows nodes to recover from crashes or network partitions. Nodes automatically detect gaps in their blockchain, identify missing epoch ranges, and request the corresponding blocks from peers using JOIN and UPDATE messages.

4. Orphan Block Management: The system includes robust handling of out-of-order block delivery. Blocks received before their parent blocks are buffered temporarily and automatically processed once the missing parent arrives, ensuring chain consistency despite network delays.

Limitations

We found no limitations with our current implementation of this project.