zeroapi 0.2.1

Zero is a RPC framework to build fast and high performance Python microservices

Zero is a RPC framework to build fast and high performance Python microservices

---

Zero is a high performance and fast (see benchmarks) Python microservice framework that provides RPC and Pub Sub interface.

Features:

• Zero provides faster communication between the microservices using zeromq under the hood.
• Zero uses messages for communication and traditional client-server or request-reply pattern is supported.
• Support for both Async and sync.
• The base server (ZeroServer) utilizes all cpu cores.
• Code generation! See example 👇

Philosophy behind Zero:

• Zero learning curve: The learning curve is tends to zero. You just add your functions and spin up a server, literally that's it! The framework hides the complexity of messaging pattern that enables faster communication.
• ZeroMQ: An awesome messaging library enables the power of Zero.

Let's get started!

Getting started 🚀

Ensure Python 3.8+

pip install zeroapi

• Create a server.py

from zero import ZeroServer

def echo(msg: str) -> msg:
    return msg

async def hello_world() -> msg:
    return "hello world"


if __name__ == "__main__":
    app = ZeroServer(port=5559)
    app.register_rpc(echo)
    app.register_rpc(hello_world)
    app.run()

Please note that server RPC methods are type hinted. Type hint is must in Zero server.

See the method type async or sync, doesn't matter. 😃

• Run it

python -m server

• Call the rpc methods

from zero import ZeroClient

zero_client = ZeroClient("localhost", 5559)

def echo():
    resp = zero_client.call("echo", "Hi there!")
    print(resp)

def hello():
    resp = zero_client.call("hello_world", None)
    print(resp)


if __name__ == "__main__":
    echo()
    hello()

Or using async client -

import asyncio

from zero import ZeroClient

zero_client = ZeroClient("localhost", 5559, use_async=True)

async def echo():
    resp = await zero_client.call_async("echo", "Hi there!")
    print(resp)

async def hello():
    resp = await zero_client.call_async("hello_world", None)
    print(resp)


if __name__ == "__main__":
    loop = asyncio.get_event_loop()
    loop.run_until_complete(echo())
    loop.run_until_complete(hello())

Code Generation! 🙌

You can also use our code generation tool to generate Python client code!

After running the server, like above, you can call the server to get the client code.

python -m zero.generate_client --host localhost --port 5559 --overwrite-dir ./my_client

It will generate client like this -

import typing  # remove this if not needed
from typing import List, Dict, Union, Optional, Tuple  # remove this if not needed
from zero import ZeroClient


zero_client = ZeroClient("localhost", 5559, use_async=False)


class RpcClient:
    def __init__(self, zero_client: ZeroClient):
        self._zero_client = zero_client

    def echo(self, msg: str) -> str:
        return self.zero_client.call("echo", msg)

    def hello_world(self, msg: str) -> str:
        return self.zero_client.call("hello_world", msg)

You can just use this -

from my_client import RpcClient, zero_client

client = RpcClient(zero_client)

if __name__ == "__main__":
    client.echo("Hi there!")
    client.hello_world(None)

Using zero.generate_client you can generate client code for even remote servers using the --host and --port options. You don't need access to the code 😃

Important notes 📝

• ZeroServer should always be run under if __name__ == "__main__":, as it uses multiprocessing.
• The methods which are under register_rpc() in ZeroServer should have type hinting, like def echo(msg: str):

Let's do some benchmarking 🤘

Zero is talking about inter service communication. In most real life scenarios, we need to call another microservice.

So we will be testing a server (gateway) calling another server for some data. (In the benchmarks folder, you can find more)

There are two endpoints in every tests,

• /hello: Just call for a hello world response 😅
• /order: Save a Order object in redis (ensure running redis-server first)

Ensure Python 3.8+

• Setup

  git clone https://github.com/Ananto30/zero.git
  cd zero
  python3 -m venv venv
  source venv/bin/activate
  pip install -r benchmarks/requirements.txt
  

• Test the current fast framework in Python (aiohttp, sanic, fastApi etc.)

  Run the gateway and server -

  sh benchmarks/asyncio/run_both.sh
  

  Run the benchmark using your favorite tool -

  $ wrk -d10s -t50 -c200 http://127.0.0.1:8000/hello
  

• Test Zero

  Run the gateway and server -

  sh benchmarks/zero/run_both.sh
  

  Run the benchmark using your favorite tool -

  $ wrk -d10s -t50 -c200 http://127.0.0.1:8000/hello
  

Compare the results! 👇

Benchmarks 🏆

Here is the result on MacBook Pro (13-inch, M1, 2020), Apple M1, 8 cores (4 performance and 4 efficiency), 8 GB RAM

For aiohttp -

> wrk -d10s -t8 -c240 http://127.0.0.1:8000/hello
Running 10s test @ http://127.0.0.1:8000/hello
  8 threads and 240 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    35.06ms   44.90ms 412.88ms   84.98%
    Req/Sec     1.56k   444.76     5.09k    77.25%
  124255 requests in 10.01s, 19.91MB read
  Socket errors: connect 0, read 102, write 0, timeout 0
Requests/sec:  12409.50
Transfer/sec:      1.99MB

> wrk -d10s -t8 -c240 http://127.0.0.1:8000/order
Running 10s test @ http://127.0.0.1:8000/order
  8 threads and 240 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    50.83ms   52.31ms 386.97ms   83.64%
    Req/Sec   774.89    219.09     1.73k    69.62%
  61807 requests in 10.03s, 14.85MB read
  Socket errors: connect 0, read 101, write 0, timeout 0
Requests/sec:   6161.43
Transfer/sec:      1.48MB

For sanic -

> wrk -d10s -t8 -c240 http://127.0.0.1:8000/hello
Running 10s test @ http://127.0.0.1:8000/hello
  8 threads and 240 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    13.69ms   14.63ms 171.07ms   87.96%
    Req/Sec     2.85k     1.00k    6.25k    76.12%
  226992 requests in 10.02s, 24.89MB read
  Socket errors: connect 0, read 141, write 0, timeout 0
Requests/sec:  22644.41
Transfer/sec:      2.48MB

> wrk -d10s -t8 -c240 http://127.0.0.1:8000/order
Running 10s test @ http://127.0.0.1:8000/order
  8 threads and 240 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    42.91ms   45.96ms 320.13ms   83.70%
    Req/Sec     0.98k   277.75     2.04k    70.96%
  77888 requests in 10.05s, 13.22MB read
  Socket errors: connect 0, read 126, write 0, timeout 0
Requests/sec:   7750.49
Transfer/sec:      1.32MB

For fastApi -

> wrk -d10s -t8 -c240 http://127.0.0.1:8000/hello
Running 10s test @ http://127.0.0.1:8000/hello
  8 threads and 240 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    30.33ms   24.27ms 215.15ms   68.17%
    Req/Sec     1.09k   348.97     2.36k    70.00%
  86908 requests in 10.04s, 11.44MB read
  Socket errors: connect 0, read 97, write 0, timeout 0
Requests/sec:   8653.16
Transfer/sec:      1.14MB

> wrk -d10s -t8 -c240 http://127.0.0.1:8000/order
Running 10s test @ http://127.0.0.1:8000/order
  8 threads and 240 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    43.16ms   27.80ms 283.60ms   68.00%
    Req/Sec   720.17    124.58     1.24k    71.62%
  57490 requests in 10.04s, 11.73MB read
  Socket errors: connect 0, read 100, write 0, timeout 0
Requests/sec:   5727.53
Transfer/sec:      1.17MB

For zero -

> wrk -d10s -t8 -c240 http://127.0.0.1:8000/hello
Running 10s test @ http://127.0.0.1:8000/hello
  8 threads and 240 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    15.10ms    9.91ms  43.16ms   47.27%
    Req/Sec     1.99k   777.55     3.64k    70.50%
  158706 requests in 10.01s, 25.43MB read
  Socket errors: connect 0, read 101, write 0, timeout 0
Requests/sec:  15853.92
Transfer/sec:      2.54MB

> wrk -d10s -t8 -c240 http://127.0.0.1:8000/order
Running 10s test @ http://127.0.0.1:8000/order
  8 threads and 240 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    24.59ms   21.15ms  94.09ms   62.77%
    Req/Sec     1.42k   614.40     8.39k    60.80%
  112807 requests in 10.10s, 27.76MB read
  Socket errors: connect 0, read 102, write 0, timeout 0
Requests/sec:  11167.89
Transfer/sec:      2.75MB

From the above numbers we can see only sanic has higher "hello world" performance than zero but again loses in the real-life example of redis order saving.

In short -

Framework	"hello world" example	redis save example
aiohttp	12,409.50 req/s	6,161.43 req/s
sanic	22,644.41 req/s	7,750.49 req/s
fastApi	8,653.16 req/s	5,727.53 req/s
zero	15,853.92 req/s	11,167.89 req/s

Todo list 📃

• Add pydantic support
• Code generation for pydantic models
• Improve error handling
• Fault tolerance

Contribution

Contributors are welcomed 🙏

Please leave a star ⭐ if you like Zero!