Friday, January 29, 2021

A Simple Multithreaded REST Service

I've talked about threading a little already.

https://beomagi.blogspot.com/2019/09/python-threading.html

https://beomagi.blogspot.com/2019/09/threading-continued-timing-thread.html

Next up is a more complete application. Lets look at a simple multithreaded web service.  Now this can be used for basic webpage services, but I think it really shines for making a REST API for quick tools.


Source:

https://github.com/beomagi/BasicRestApp/blob/main/http_server_mt3.py

snapshot below:

'''    Basic threaded http server implementation    '''
#!/usr/bin/env python3
import socketserver
import http.server
import sys
import time
import urllib
import json

#-------------sample functions--------------------
keyval={}
def kput(a,b):
    keyval[a]=b
    return json.dumps({a:b})

def kget(a):
    return keyval.get(a)

def add2numbers(a,b):
    return int(a)+int(b)

def handlejob(func_and_params):
    func=func_and_params[0]
    params=func_and_params[1:]
    if func=="add"return add2numbers(params[0],params[1])
    if func=="put"return kput(params[0],params[1])
    if func=="get"return kget(params[0])
    return "Unknown Command"
#-^-^--^-^-^-^sample functions-^-^-^-^-^-^-^-^-^-^-


class Handler(http.server.BaseHTTPRequestHandler):
    '''   use our own handlers functions '''

    def sendtextinfo(selfcodetext):
        self.send_response(code)
        self.send_header('Content-type''text/html')
        self.end_headers()
        if type(text)==type([]):
            for lines in text:
                self.wfile.write((str(lines)+"\n").encode())
        else:
            self.wfile.write((str(text)+"\n").encode())

    def do_GET(self):
        '''   handle get   '''
        tnow = time.time()
        gnow = time.strftime('%Y-%m-%d %H:%M:%S', time.gmtime(tnow)) #Formatted UTC time
        parsed_data = urllib.parse.urlparse(self.path)
        if parsed_data.geturl().lower() == "/time":
            message = gnow
        else : message="this is a test of the multithreaded webservice"
        self.sendtextinfo(200,message)

    def do_POST(self):
        '''   handle post like rest API   '''
        try#try getting the bytestream of the request
            content_length = int(self.headers['Content-Length'])
        except Exception as err:
            print("malformed headers")
            self.sendtextinfo(200,str(err))
            return

        if content_length > 0:
            rawrequest = self.rfile.read(content_length).decode('utf-8')
            print("Received POST: {}".format(rawrequest))
            try:
                jrequest = json.loads(rawrequest)
            except BaseException as anError:
                self.sendtextinfo(200,"Error in JSON: {}".format(str(anError)))
                return

        if "cmd" in jrequest:
            commandandparams=jrequest['cmd']
            print("Command received: {}".format(commandandparams))
            result=handlejob(commandandparams)
            if result != None:
                self.sendtextinfo(200,result)
            else:
                self.sendtextinfo(200,"No Output")
        return


class ThreadedHTTPServer(socketserver.ThreadingMixInhttp.server.HTTPServer):
    '''    Basic threaded server class    '''
    http.server.HTTPServer.request_queue_size = 128

if sys.argv[1:]:
    HTPORT = int(sys.argv[1])
else:
    HTPORT = 8000

HTSERVER = ThreadedHTTPServer((''HTPORT), Handler)

try:
    while 1:
        sys.stdout.flush()
        HTSERVER.handle_request()
except KeyboardInterrupt:
    print("Server Stopped")


Discussion:

I've highlighted different sections of the code by altering the background of the div.

The bottom section with the dark purple background is our main area - could stick this in a construct like "if __name__ == '__main__':" but it's not necessary for our purposes.

The main here does a check for passed parameters to override the default port the application will be served on. Then we have an infinite loop that keep calling handle_request(). The "flush" before that forces any buffering in the io to be handled before the next handle. Useful for problems where buffered output isn't yet written at the time of a crash.

The main section makes a call to ThreadedHTTPServer. This manages threaded calls to the handler(3rd section). The handler implements calls for GET and POST. 

I've added some simple functions on POST calls for adding 2 numbers, storing, and getting values in a key-value store sort of setup.

Interfacing with POST can be done from any other language. It's also nice to use CURL to test REST interfaces.



e.g.

beomagi@BeoBalthazar ~/gits/beomagi/BasicRestApp  (main) 2021-01-29 13:35:09

└─ $ ∙ curl 127.0.0.1:8000

this is a test of the multithreaded webservice

beomagi@BeoBalthazar ~/gits/beomagi/BasicRestApp  (main) 2021-01-29 13:35:16

└─ $ ∙ curl 127.0.0.1:8000/time

2021-01-29 18:35:21

beomagi@BeoBalthazar ~/gits/beomagi/BasicRestApp  (main) 2021-01-29 13:35:16

└─ $ ∙ curl 127.0.0.1:8000 -X POST --data '{"cmd":["add",51,15]}'

66


So this example shows how we can make a processing request to the server. Obviously this can be more complex. Maybe I'm sending a url to be parsed (get me ma comix!). Maybe I'm passing data I want to store. Numerous reasons for this. What's performance like?


beomagi@BeoBalthazar ~/gits/beomagi/BasicRestApp  (main) 2021-01-29 13:35:16

└─ $ ∙ curl 127.0.0.1:8000 -X POST --data '{"cmd":["put","b","If A Technological Feat Is Possible, Man Will Do It. Almost As If It is Wired Into The Core Of Our Being."]}'

{"b": "If A Technological Feat Is Possible, Man Will Do It. Almost As If It is Wired Into The Core Of Our Being."}

beomagi@BeoBalthazar ~/gits/beomagi/BasicRestApp  (main) 2021-01-29 13:35:16

└─ $ ∙ curl 127.0.0.1:8000 -X POST --data '{"cmd":["get","b"]}'

If A Technological Feat Is Possible, Man Will Do It. Almost As If It is Wired Into The Core Of Our Being.



Ok, so that's the get and put app level functions working. Let's try a bunch of gets.


beomagi@BeoBalthazar ~/gits/beomagi/BasicRestApp  (main) 2021-01-29 13:35:16

└─ $ ∙ time (for a in {1..1000}; do (curl -s 127.0.0.1:8000 -X POST --data '{"cmd":["get","b"]}' )  & done |sort | uniq ; wait )

If A Technological Feat Is Possible, Man Will Do It. Almost As If It is Wired Into The Core Of Our Being.


real    0m2.403s

user    0m4.234s

sys     0m20.688s

beomagi@BeoBalthazar ~/gits/beomagi/BasicRestApp  (main) 2021-01-29 13:35:16

└─ $ ∙


The time it took is the real value. It took 2.4 seconds to pull this data a thousand times. If you want to pull a lot of data faster, this basic server can be expanded to use multi-put/multi-get

This isn't great though. We're severely limiting the test by relying on bash to loop though and spawn requests. Bash is slow for this :)

So let's try apache bench.

Start by putting the request in a file.

beomagi@BeoBalthazar ~/gits/beomagi/BasicRestApp  (main) 2021-01-29 13:35:16

└─ $ ∙ cat postthis.txt

'{"cmd":["get","b"]}'


Now lets run apache bench to stress this.

beomagi@BeoBalthazar ~/gits/beomagi/BasicRestApp  (main) 2021-01-29 13:35:16

└─ $ ∙ ab -n 10000 -c 128  -p ./postthis.txt -T application/json http://127.0.0.1:8000/
This is ApacheBench, Version 2.3 <$Revision: 1807734 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking 127.0.0.1 (be patient)
Completed 1000 requests
Completed 2000 requests
Completed 3000 requests
Completed 4000 requests
Completed 5000 requests
Completed 6000 requests
Completed 7000 requests
Completed 8000 requests
Completed 9000 requests
Completed 10000 requests
Finished 10000 requests


Server Software:        BaseHTTP/0.6
Server Hostname:        127.0.0.1
Server Port:            8000

Document Path:          /
Document Length:        57 bytes

Concurrency Level:      128
Time taken for tests:   3.378 seconds
Complete requests:      10000
Failed requests:        0
Total transferred:      1730000 bytes
Total body sent:        1570000
HTML transferred:       570000 bytes
Requests per second:    2960.29 [#/sec] (mean)
Time per request:       43.239 [ms] (mean)
Time per request:       0.338 [ms] (mean, across all concurrent requests)
Transfer rate:          500.13 [Kbytes/sec] received
                        453.87 kb/s sent
                        954.00 kb/s total

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    2   5.1      0      47
Processing:     5   40  10.6     41      85
Waiting:        5   38   9.8     40      67
Total:         12   43   9.5     41     100

Percentage of the requests served within a certain time (ms)
  50%     41
  66%     44
  75%     47
  80%     48
  90%     55
  95%     61
  98%     68
  99%     73
 100%    100 (longest request)

_____________________________________________________________


Using apachebench (ab) loading the queue to the max 128 concurrent requests, we handle 10k requests in 3.378 seconds.


I've used this in another system where getting outside software was quite restrictive. It was quicker to write something like the above to act as a cache for a monitoring system. The system would store various database counts and system checks for a monitoring system. Prior to that the monitoring system was making the DB queries directly and wow did our DB hate that!!. Granted it was far more fleshed out - there was get/put single value and multiple value commands. The default page would show all stored data, and the age of the data. There was a timing thread for stats on how heavy the system was being hit. While that system (cachewho) was written and made for that company, I do plan to rewrite in python3 (instead of 2).

Saturday, January 16, 2021

Virus Infection Simulator

Was talking to my S-I-L about some school projects and an idea for this came up. Out of scope from what her project required, but in scope for something to do on a day off!

Direct link to github html preview.

I started by defining a list of "cities" - x, y coordinates are simply distributed in a circle around the center of the screen.

People is an array of persons. A person is in a city (thus having that city's x, y location as a base point. The person wanders based on sin/cos curves.

e.g. lets say the person wanders around the city. Cities are arranged in a circle. we can describe a function given a time to determine the location of a person as they wander the city.

It's  not hard to think this through. A "person" will need some stats. Direction, speed, wandering constants, home city. "Cities" will have constants too - size, location. We can consider population density. Maybe some cities have mandates that reduce infection rate.

//Basic wandering idea... 

//lets give a person a random direction, clockwise or anti

persondirection=random()-0.5

//wander factor is how often the person will head to the center of a town while they run around.

wanderfactor=random()*4+2

//Given a number of cities, we can distribute cities in a circle.

cityx=sin(2*pi*personcitynumber/citycount)*cityspread

cityy=cos(2*pi*personcitynumber/citycount)*cityspread

//lets have the person walk in a circle in the direction/speed of their random pick. Time is a variable we can increment every frame. Offset is a random constant per person. It's so that that everyone doesn't seem to start at the same point at the same time.

wanderingx=sin(time*persondirection+offset)*citysize

wanderingy=cos(time*persondirection+offset)*citysize

//we can make the person wander a little more than just in a perfect circle. Sin/2 is -0.5 to 0.5. Add 0.5 and it's 0 to 1. 

centervisit=0.5+sin(time*wanderfactor+offset)/2

//add the city coordinates with the wandering x,y multiplied by the wander factor, and they'll be moving around a city location, moving to the center and back out to their normal ring around the city.

personx cityx centervisit wanderingx

persony cityy centervisit wanderingy

Fleshed out a little more, the wandering patterns looked pretty decent. Throw these values in an object per person, make an array of people, and city variables, and add some more properties like what city the person is moving to, and progress there.

See the source here

https://github.com/beomagi/html-tinkers/blob/master/virus-sim/virus-sim.html

Edit: added some color, event log and graph. See the git repo for future changes.

Thought: should make this an interactive game. Cities can have money saved that decreases if under lockdown. People can have a happiness factor that depends on safeguards and how they are doing in relation to other cities...