Service Streamer

Boosting your Web Services of Deep Learning Applications. 中文README

What is Service Streamer ? • Highlights • Installation • Develop BERT Service in 5 Minutes • API • Benchmark • FAQ •

• Made by ShannonAI • 🌐 http://www.shannonai.com/

What is Service Streamer ?

A mini-batch collects data samples and is usually used in deep learning models. In this way, models can utilize the parallel computing capability of GPUs. However, requests from users for web services are usually discrete. If using conventional loop server or threaded server, GPUs will be idle dealing with one request at a time. And the latency time will be linearly increasing when there are concurrent user requests.

ServiceStreamer is a middleware for web service of machine learning applications. Queue requests from users are sampled into mini-batches. ServiceStreamer can significantly enhance the overall performance of the system by improving GPU utilization.

Highlights

🐣 Easy to use: Minor changes can speed up the model ten times.
⚡ Fast processing speed: Low latency for online inference of machine learning models.
🐙 Good expandability: Easy to be applied to multi-GPU scenarios for handling enormous requests.
⚔️ Applicability: Used with any web frameworks and/or deep learning frameworks.

Installation

Install ServiceStream by using pip，requires Python >= 3.5 :

pip install service_streamer

Develop BERT Service in 5 Minutes

We provide a step-by-step tutorial for you to bring BERT online in 5 minutes. The service processes 1400 sentences per second.

Text Infilling is a task in natural language processing: given a sentence with several words randomly removed, the model predicts those words removed through the given context.

BERT has attracted a lot of attention in these two years and it achieves State-Of-The-Art results across many nlp tasks. BERT utilizes "Masked Language Model (MLM)" as one of the pre-training objectives. MLM models randomly mask some of the tokens from the input, and the objective is to predict the original vocabulary id of the masked word based on its context. MLM has similarities with text infilling. It is natural to introduce BERT to text infilling task.

First, we define a model for text filling task bert_model.py. The predict function accepts a batch of sentences and returns predicted position results of the [MASK] token.

class TextInfillingModel(object):
    ...


batch = ["twinkle twinkle [MASK] star.",
         "Happy birthday to [MASK].",
         'the answer to life, the [MASK], and everything.']
model = TextInfillingModel()
outputs = model.predict(batch)
print(outputs)
# ['little', 'you', 'universe']

Note: Please download pre-trained BERT model at first.

Second, utilize Flask to pack predicting interfaces to Web service. flask_example.py

model = TextInfillingModel()
@app.route("/naive", methods=["POST"])
def naive_predict():
    inputs = request.form.getlist("s")
    outputs = model.predict(inputs)
    return jsonify(outputs)
 
app.run(port=5005)

Please run flask_example.py, then you will get a vanilla Web server.

curl -X POST http://localhost:5005/naive -d 's=Happy birthday to [MASK].' 
["you"]

At this time, your web server can only serve 12 requests per second. Please see benchmark for more details.

Third, encapsulate model functions through service_streamer. Three lines of code make the prediction speed of BERT service reach 200+ sentences per second (16x faster).

from service_streamer import ThreadedStreamer
streamer = ThreadedStreamer(model.predict, batch_size=64, max_latency=0.1)

@app.route("/stream", methods=["POST"])
def stream_predict():
    inputs = request.form.getlist("s")
    outputs = streamer.predict(inputs)
    return jsonify(outputs)

app.run(port=5005, debug=False)

Run flask_example.py and test the performance with wrk.

wrk -t 2 -c 128 -d 20s --timeout=10s -s benchmark.lua http://127.0.0.1:5005/stream
...
Requests/sec:    200.31

Finally, encapsulate models through Streamer and start service workers on multiple GPUs. Streamer further accelerates inference speed and achieves 1000+ sentences per second (80x faster).

from service_streamer import ManagedModel, Streamer

class ManagedBertModel(ManagedModel):

    def init_model(self):
        self.model = TextInfillingModel()

    def predict(self, batch):
        return self.model.predict(batch)

streamer = Streamer(ManagedBertModel, batch_size=64, max_latency=0.1, worker_num=8, cuda_devices=(0, 1, 2, 3))
app.run(port=5005, debug=False)

8 gpu workers can be started and evenly distributed on 4 GPUs.

API

Quick Start

In general, the inference speed will be faster by utilizing parallel computing.

outputs = model.predict(batch_inputs)

ServiceStreamer is a middleware for web service of machine learning applications. Queue requests from users are scheduled into mini-batches and forward into GPU workers. ServiceStreamer sacrifices a certain delay (default maximum is 0.1s) and enhance the overall performance by improving the ratio of GPU utilization.

from service_streamer import ThreadedStreamer

# Encapsulate batch_predict function with Streamer

streamer = ThreadedStreamer(model.predict, batch_size=64, max_latency=0.1)

# Replace model.predict with streamer.predict

outputs = streamer.predict(batch_inputs)

Start web server on multi-threading (or coordination). Your server can usually achieve 10x (batch_size/batch_per_request) times faster by adding a few lines of code.

Distributed GPU worker

The performance of web server (QPS) in practice is much higher than that of GPU model. We also support one web server with multiple GPU worker processes.

from service_streamer import Streamer

# Spawn releases 4 gpu worker processes
streamer = Streamer(model.predict, 64, 0.1, worker_num=4)
outputs = streamer.predict(batch)

Streamer uses spawn subprocesses to run gpu workers by default. Streamer uses interprocess queues to communicate and queue. It can distribute a large number of requests to multiple workers for processing.

Then the prediction results of the model are returned to the corresponding web server in batches. And results are forwarded to the corresponding http response.

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 390.116                Driver Version: 390.116                   |
|-------------------------------+----------------------+----------------------+
...
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      7574      C   /home/liuxin/nlp/venv/bin/python            1889MiB |
|    1      7575      C   /home/liuxin/nlp/venv/bin/python            1889MiB |
|    2      7576      C   /home/liuxin/nlp/venv/bin/python            1889MiB |
|    3      7577      C   /home/liuxin/nlp/venv/bin/python            1889MiB |
+-----------------------------------------------------------------------------+

The above method is simple to define, but the main process initialization model takes up an extra portion of memory. And the model can only run on the same GPU. Therefore, we have provided the ManagedModel class to facilitate model lazy initialization and migration while supporting multiple GPUs.

from service_streamer import ManagedModel

class ManagedBertModel(ManagedModel):

    def init_model(self):
        self.model = Model()

    def predict(self, batch):
        return self.model.predict(batch)


# Spawn produces 4 gpu worker processes, which are evenly distributed on 0/1/2/3 GPU
streamer = Streamer(ManagedBertModel, 64, 0.1, worker_num=4, cuda_devices=(0, 1, 2, 3))
outputs = streamer.predict(batch)

Distributed Web Server

Some cpu-intensive calculations, such as image and text preprocessing, need to be done first in web server. The preprocessed data is then forward into GPU worker for predictions. CPU resources often become performance bottlenecks in practice. Therefore, we also provide the mode of multi-web servers matching (single or multiple) gpu workers.

Use RedisStream to specify a unique Redis address for all web servers and gpu workers.

# default parameters can be omitted and localhost:6379 is used.
streamer = RedisStreamer(redis_broker="172.22.22.22:6379")

We make use of gunicorn or uwsgi to implement reverse proxy and load balancing.

cd example
gunicorn -c redis_streamer_gunicorn.py flask_example:app

Each request will be load balanced to each web server for cpu preprocessing, and then evenly distributed to gpu worker for model prediction.

Future API

You might be familiar with future if you have used any concurrent library. You can use the Future API directly if you want to use service_streamer for queueing requests or distributed GPU computing and using scenario is not web service.

from service_streamer import ThreadedStreamer
streamer = ThreadedStreamer(model.predict, 64, 0.1)

xs = []
for i in range(200):
    future = streamer.submit([["How", "are", "you", "?"], ["Fine", "."], ["Thank", "you", "."]])
    xs.append(future)


# Get all instances of future object and wait for asynchronous responses. 
for future in xs:
    outputs = future.result()
    print(outputs)

Benchmark

We utilize wrk to conduct benchmark test.

Test examples and scripts can be found in example.

Environment

gpu: Titan Xp
cuda: 9.0
pytorch: 1.1

Single GPU process

# start flask threaded server
python example/flask_example.py

# benchmark naive api without service_streamer
wrk -t 4 -c 128 -d 20s --timeout=10s -s benchmark.lua http://127.0.0.1:5005/naive
# benchmark stream api with service_streamer
wrk -t 4 -c 128 -d 20s --timeout=10s -s benchmark.lua http://127.0.0.1:5005/stream

	Naive	ThreaedStreamer	Streamer	RedisStreamer
qps	12.78	207.59	321.70	372.45
latency	8440ms	603.35ms	392.66ms	340.74ms

Multiple GPU processes

The performance loss of the communications and load balancing mechanism of multi-gpu workers are verified compared with a single web server process.

We adopt gevent server because multi-threaded Flask server has become a performance bottleneck. Please refer to the flask_multigpu_example.py

wrk -t 8 -c 512 -d 20s --timeout=10s -s benchmark.lua http://127.0.0.1:5005/stream

gpu_worker_num	Naive	ThreadedStreamer	Streamer	RedisStreamer
1	11.62	211.02	362.69	365.80
2	N/A	N/A	488.40	609.63
4	N/A	N/A	494.20	1034.57

Threaded Streamer Due to the limitation of Python GIL, multi-worker is meaningless. We conduct comparison studies using single GPU worker.
Streamer Performance improvement is not linear when it is greater than 2 gpu worker. The utilization rate of CPU reaches 100. The bottleneck is CPU at this time and the performance issue of flask is the obstacle.

Utilize Future API to start multiple GPU processes

We adopt Future API to conduct multi-GPU benchmeark test locally in order to reduce the performance influence of web server. Please refer to code example in future_example.py

gpu_worker_num	Batched	ThreadedStreamer	Streamer	RedisStreamer
1	422.883	401.01	399.26	384.79
2	N/A	N/A	742.16	714.781
4	N/A	N/A	1400.12	1356.47

It can be seen that the performance of service_streamer is almost linearly related to the number of gpu workers. Communications of inter-process in service_streamer is more efficient than redis.

FAQ

Q: using a model trained from allennlp,set worker_num=4 of Streamer during inference, what's the reason that 16-core cpu is full and speed is slower than Streamer with worker_num=1?

A: for multi-process inference, if the model process data using numpy with multi-thread, it may cause cpu overheads, resulting in a multi-core computing speed that slower than a single core. This kind of problem may occur when using third-party libraries such as alennlp, spacy, etc. It could be solved by setting numpy threadsenvironment variables.

import os
os.environ["MKL_NUM_THREADS"] = "1"  # export MKL_NUM_THREADS=1 
os.environ["NUMEXPR_NUM_THREADS"] = "1"  # export NUMEXPR_NUM_THREADS=1 
os.environ["OMP_NUM_THREADS"] = "1"  # export OMP_NUM_THREADS=1
import numpy

make sure putting environment variables before import numpy.

Q: When using RedisStreamer, if there are only one redis broker and more than one model, the input batches may have different structure. How to deal with such situation?

A: Specify the prefix when initializing worker and streamer, each streamer will use a unique channel.

example of initialiazing workers:

from service_streamer import run_redis_workers_forever
from bert_model import ManagedBertModel

if __name__ == "__main__":
    from multiprocessing import freeze_support
    freeze_support()
    run_redis_workers_forever(ManagedBertModel, 64, prefix='channel_1')
    run_redis_workers_forever(ManagedBertModel, 64, prefix='channel_2')

example of using streamer to have result:

from service_streamer import RedisStreamer

streamer_1 = RedisStreaemr(prefix='channel_1')
streamer_2 = RedisStreaemr(prefix='channel_2')

# predict
output_1 = streamer_1.predict(batch)
output_2 = streamer_2.predict(batch)

ShannonAI/service-streamer

Service Streamer

• Made by ShannonAI • 🌐 http://www.shannonai.com/

What is Service Streamer ?

Highlights

Installation

Develop BERT Service in 5 Minutes

API

Quick Start

Distributed GPU worker

Distributed Web Server

Future API

Benchmark

Benchmark

Environment

Single GPU process

Multiple GPU processes

Utilize Future API to start multiple GPU processes

FAQ