How vpatel95/hyper-line works
Main Components
a. User Node (avd_user)
• Loads the user’s application as a shared object (.so)
• Reads user.json config
• Initiates pipeline request to Server over TCP
b. Server Node (avd_server)
• Reads server.json config
• Listens on two ports: one for user requests, one for worker registrations
• Manages task orchestration, worker registry, and final result collation
c. Worker Nodes (avd_worker)
• Each runs on its own machine
• Reads worker.json config
• Registers with Server
• Hosts a peer‐to‐peer server for inter‐stage communication
• Executes one pipeline stage (invokes user .so function)
d. Shared Libraries & Tools
• avd_log (logging)
• cJSON (JSON parsing)
• avd_pipe (message‐passing primitives)
• src/tools/avd_sim (optional simulator)
Relationships & Data Flows
1. User → Server: Pipeline request + serialized user config
2. Server → Worker: Stage assignment + input data
3. Worker → Worker: Stage‐to‐stage data forwarding (peer ports)
4. Last Worker → Server: Final stage output
5. Server → User: Aggregated result
Architectural Patterns
• Distributed assembly‐line/pipeline pattern
• Plugin pattern (user code loaded as .so)
• Master–worker orchestration (Server is Master, Workers are Slaves)
• Peer‐to‐peer chaining between Workers
Call out Key Technologies & Files
• Makefile/Common.mk: build orchestration
• src/apps/...: sample applications (encryption.so, sum.so)
• src/lib: core libraries (avd_log.c, cJSON.c)
• src/include: public APIs (avd_session.h, avd_message.h, etc.)
• src/server, src/worker, src/user: binaries avd_server, avd_worker, avd_user
Detailed Elements
• Clear labels on each component: name, binary filename, listening ports