Worker-as-service is a high-efficient deployment framework for general computing:
- Low latency & high throughput messaging pipeline
- Micro-batching
- Ease of use, event data science can deploy a service with low maintanance effort
- Decentralize deployment
- Native python support, pre-defined protobuf (gRPC likes) is needed
- Resful API support
- Statistic support out-of-the-box
-
step 1: clone this project to local machine
-
step 2: navigate to
worker-as-service/server
:
pip install -e .
- step 3: navigate to
worker-as-service/client
:
pip install -e .
- step 4: Checkout example in
worker-as-service/example
for custom server:
- Clip-as-service (formerly known as Bert-as-service): Worker-as-service is heavily based on Bert-as-service.
- PyZMQ: for effience transport layer.