We Keep Coding

Multiprocessing Object Method with Pathos Has no Speed Up

I want to call the buildWorld method of each of several world objects. This is a computationally expensive and time-consuming task, sometimes taking several hours depending on the settings you give each world.
After initially running into the issue that pickle couldn't serialize my object methods (described here and in several questions like here and here), I tried using pathos. The multiprocessing did not improve performance, so I did a time check on just pure serialization of my objects to see if this was the source of the slowdown, but it wasn't. Nevertheless, I eventually took an approach where I created a world within the child process rather than pass it in, then saved it as a file that I could re-load in my parent process to avoid having to serialize anything on a return (and also on child process start, I assume?). I came across several similar questions (A & B), and tried a few things (see below), but none of the answers seemed to do the trick for me.
Unfortunately, nothing actually sped up my code. I was seeing that child processes were being created, but if I created 3 subprocesses, for example, my code ran 3x slower, and so took just as much time as if I ran just one process. Any guidance is appreciated!
Method 1 (Pathos ProcessingPool):
import pathos.multiprocessing as mp
def run_world_building(worldNum):
myWorld = emptyWorld(worldNum) # not expensive
myWorld.buildWorld() # very expensive
myWorld.save() # create a file with world info
p = mp.ProcessingPool(3)
p.map(run_world_building, range(0,3))
Method 2 (Multiprocess (no '-ing') with individual Process objects):
import multiprocess as mp
def run_world_building(worldNum):
myWorld = emptyWorld(worldNum) # not expensive
myWorld.buildWorld() # very expensive
myWorld.save() # create a file with world info
processes = []
for i in range(0,3):
p = mp.Process(target=run_world_building, args=(i,))
processes.append(p)
# I separated the start and join loops, but
# not sure if that's entirely necessary
for i in range(0, 3):
processes[i].start()
for i in range(0,3):
processes[i].join()
Method 3 (Using ThreadPool):
from pathos.pools import ThreadPool
def run_world_building(worldNum):
myWorld = emptyWorld(worldNum) # not expensive
myWorld.buildWorld() # very expensive
myWorld.save() # create a file with world info
p = ThreadPool(3)
p.map(run_world_building, range(0,3))

Multiprocessing where new process starts hafway through other process

I have a Python script that does two things; 1) it downloads a large file by making an API call, and 2) preprocess that large file. I want to use Multiprocessing to run my script. Each individual part (1 and 2) takes quite long. Everything happens in-memory due to the large size of the files, so ideally a single core would do both (1) and (2) consecutively. I have a large amount of cores available (100+), but I can only have 4 API calls running at the same time (limitation set by the API developers). So what I want to do is spawn 4 cores that start downloading by making an API-call, and as soon as one of those cores is done downloading and starts preprocessing I want a new core to start the whole process as well. This so there's always 4 cores downloading, and as many cores as needed doing the pre-processing. I do not know however how to have a new core spawn as soon as another core is finished with the first part of the script.
My actual code is way too complex to just dump here, but let's say I have the following two functions:
import requests
def make_api_call(val):
"""Function that does part 1); makes an API call, stores it in memory and returns a large
satellite GeoTIFF
"""
large_image = requests.get(val)
return(large_image)
def preprocess_large_image(large_image):
"""Function that does part 2); preprocesses a large image, and returns the relevant data
"""
results = preprocess(large_image)
return(results)
how then can I make sure that as soon as a single core/process is finished with 'make_api_call' and starts with 'preprocess_large_image', another core spawns and starts the entire process as well? This so there is always 4 images downloading side-by-side. Thank you in advance for the help!

This is a perfect application for a multiprocessing.Semaphore (or for safety, use a BoundedSemaphore)! Basically you put a lock around the api call part of the process, but let up to 4 worker processes hold the lock at any given time. For various reasons, things like Lock, Semaphore, Queue, etc all need to be passed at the creation of a Pool, rather than when a method like map or imap is called. This is done by specifying an initialization function in the pool constructor.
def api_call(arg):
return foo
def process_data(foo):
return "done"
def map_func(arg):
global semaphore
with semaphore:
foo = api_call(arg)
return process_data(foo)
def init_pool(s):
global semaphore = s
if __name__ == "__main__":
s = mp.BoundedSemaphore(4) #max concurrent API calls
with mp.Pool(n_workers, init_pool, (s,)) as p: #n_workers should be great enough that you always have a free worker waiting on semaphore.acquire()
for result in p.imap(map_func, arglist):
print(result)

If both the downloading (part 1) and the conversion (part 2) take long, there is not much reason to do everything in memory.
Keep in mind that networking is generally slower than disk operations.
So I would suggest to use two pools, saving the downloaded files to disk, and send file names to workers.
The first Pool is created with four workers and does the downloading. The worker saves the image to a file and returns the filename. With this Pool you use the imap_unordered method, because that starts yielding values as soon as they become available.
The second Pool does the image processing. It gets fed by apply_async, which returns an AsyncResult object.
We need to save those to keep track of when all the conversions are finished.
Note that map or imap_unordered are not suitable here because they require a ready-made iterable.
def download(url):
large_image = requests.get(url)
filename = url_to_filename(url) # you need to write this
with open(filename, "wb") as imgf:
imgf.write(large_image)
def process_image(name):
with open(name, "rb") as f:
large_image = f.read()
# File processing goes here
with open(name, "wb") as f:
f.write(large_image)
return name
dlp = multiprocessing.Pool(processes=4)
# Default pool size is os.cpu_count(); might be too much.
imgp = multiprocessing.Pool(processes=20)
urllist = ['http://foo', 'http://bar'] # et cetera
in_progress = []
for name in dlp.imap_unordered(download, urllist):
in_progress.append(imgp.apply_async(process_image, (name,)), )
# Wait for the conversions to finish.
while in_progress:
finished = []
for res in in_progress:
if res.ready():
finished.append(res)
for f in finished:
in_progress.remove(f)
print(f"Finished processing '{f.get()}'.")
time.sleep(0.1)

Python asyncio: Queue concurrent to normal code

Edit: I am closing this question.
As it turns out, my goal of having parallel HTTP posts is pointless. After implementing it successfully with aiohttp, I run into deadlocks elsewhere in the pipeline.
I will reformulate this and post a single question in a few days.
Problem
I want to have a class that, during some other computation, holds generated data and can write it to a DB via HTTP (details below) when convenient. It's gotta be a class as it is also used to load/represent/manipulate data.
I have written a naive, nonconcurrent implementation that works:
The class is initialized and then used in a "main loop". Data is added to it in this main loop to a naive "Queue" (a list of HTTP requests). At certain intervals in the main loop, the class calls a function to write those requests and clear the "queue".
As you can expect, this is IO bound. Whenever I need to write the "queue", the main loop halts. Furthermore, since the main computation runs on a GPU, the loop is also not really CPU bound.
Essentially, I want to have a queue, and, say, ten workers running in the background and pushing items to the http connector, waiting for the push to finish and then taking on the next (or just waiting for the next write call, not a big deal). In the meantime, my main loop runs and adds to the queue.
Program example
My naive program looks something like this
class data_storage(...):
def add(...):
def write_queue(self):
if len(self.queue) > 0:
res = self.connector.run(self.queue)
self.queue = []
def main_loop(storage):
# do many things
for batch in dataset: #simplified example
# Do stuff
for some_other_loop:
(...)
storage.add(results)
# For example, call each iteration
storage.write_queue()
if __name__ == "__main__":
storage=data_storage()
main_loop(storage)
...
In detail: the connector class is from the package 'neo4j-connector' to post to my Neo4j database. It essentially does JSON formatting and uses the "requests" api from python.
This works, even without a real queue, since nothing is concurrent.
Now I have to make it work concurrently.
From my research, I have seen that ideally I would want a "producer-consumer" pattern, where both are initialized via asyncio. I have only seen this implemented via functions, not classes, so I don't know how to approach this. With functions, my main loop should be a producer coroutine and my write function becomes the consumer. Both are initiated as tasks on the queue and then gathered, where I'd initialize only one producer but many consumers.
My issue is that the main loop includes parts that are already parallel (e.g. PyTorch). Asyncio is not thread safe, so I don't think I can just wrap everything in async decorators and make a co-routine. This is also precisely why I want the DB logic in a separate class.
I also don't actually want or need the main loop to run "concurrently" on the same thread with the workers. But it's fine if that's the outcome as the workers don't do much on the CPU. But technically speaking, I want multi-threading? I have no idea.
My only other option would be to write into the queue until it is "full", halt the loop and then use multiple threads to dump it to the DB. Still, this would be much slower than doing it while the main loop is running. My gain would be minimal, just concurrency while working through the queue. I'd settle for it if need be.
However, from a stackoverflow post, I came up with this small change
class data_storage(...):
def add(...):
def background(f):
def wrapped(*args, **kwargs):
return asyncio.get_event_loop().run_in_executor(None, f, *args, **kwargs)
return wrapped
#background
def write_queue(self):
if len(self.queue) > 0:
res = self.connector.run(self.queue)
self.queue = []
Shockingly this sort of "works" and is blazingly fast. Of course since it's not a real queue, things get overwritten. Furthermore, this overwhelms or deadlocks the HTTP API and in general produces a load of errors.
But since this - in principle - works, I wonder if I could do is the following:
class data_storage(...):
def add(...):
def background(f):
def wrapped(*args, **kwargs):
return asyncio.get_event_loop().run_in_executor(None, f, *args, **kwargs)
return wrapped
#background
def post(self, items):
if len(items) > 0:
self.nr_workers.increase()
res = self.connector.run(items)
self.nr_workers.decrease()
def write_queue(self):
if self.nr_workers < 10:
items=self.queue.get(200) # Extract and delete from queue, non-concurrent
self.post(items) # Add "Worker"
for some hypothetical queue and nr_workers objects. Then at the end of the main loop, have a function that blocks progress until number of workers is zero and clears, non-concurrently, the rest of the queue.
This seems like a monumentally bad idea, but I don't know how else to implement this. If this is terrible, I'd like to know before I start doing more work on this. Do you think it would it work?
Otherwise, could you give me any pointers as how to approach this situation correctly?
Some key words, tools or things to research would of course be enough.
Thank you!

multithreading check membership in Queue and stop the threads

I want to iterate over a list using 2 thread. One from leading and other from trailing, and put the elements in a Queue on each iteration. But before putting the value in Queue I need to check for existence of the value within Queue (its when that one of the threads has putted that value in Queue), So when this happens I need to stop the thread and return list of traversed values for each thread.
This is what I have tried so far :
from Queue import Queue
from threading import Thread, Event
class ThreadWithReturnValue(Thread):
def __init__(self, group=None, target=None, name=None,
args=(), kwargs={}, Verbose=None):
Thread.__init__(self, group, target, name, args, kwargs, Verbose)
self._return = None
def run(self):
if self._Thread__target is not None:
self._return = self._Thread__target(*self._Thread__args,
**self._Thread__kwargs)
def join(self):
Thread.join(self)
return self._return
main_path = Queue()
def is_in_queue(x, q):
with q.mutex:
return x in q.queue
def a(main_path,g,l=[]):
for i in g:
l.append(i)
print 'a'
if is_in_queue(i,main_path):
return l
main_path.put(i)
def b(main_path,g,l=[]):
for i in g:
l.append(i)
print 'b'
if is_in_queue(i,main_path):
return l
main_path.put(i)
g=['a','b','c','d','e','f','g','h','i','j','k','l']
t1 = ThreadWithReturnValue(target=a, args=(main_path,g))
t2 = ThreadWithReturnValue(target=b, args=(main_path,g[::-1]))
t2.start()
t1.start()
# Wait for all produced items to be consumed
print main_path.join()
I used ThreadWithReturnValue that will create a custom thread that returns the value.
And for membership checking I used the following function :
def is_in_queue(x, q):
with q.mutex:
return x in q.queue
Now if I first start the t1 and then the t2 I will get 12 a then one b then it doesn't do any thing and I need to terminate the python manually!
But if I first run the t2 then t1 I will get the following result:
b
b
b
b
ab
ab
b
b
b
b
a
a
So my questions is that why python treads different in this cases? and how can I terminate the threads and make them communicate with each other?

Before we get into bigger problems, you're not using Queue.join right.
The whole point of this function is that a producer who adds a bunch of items to a queue can wait until the consumer or consumers have finished working on all of those items. This works by having the consumer call task_done after they finish working on each item that they pulled off with get. Once there have been as many task_done calls as put calls, the queue is done. You're not doing a get anywhere, much less a task_done, so there's no way the queue can ever be finished. So, that's why you block forever after the two threads finish.
The first problem here is that your threads are doing almost no work outside of the actual synchronization. If the only thing they do is fight over a queue, only one of them is going to be able to run at a time.
Of course that's common in toy problems, but you have to think through your real problem:
If you're doing a lot of I/O work (listening on sockets, waiting for user input, etc.), threads work great.
If you're doing a lot of CPU work (calculating primes), threads don't work in Python because of the GIL, but processes do.
If you're actually primarily dealing with synchronizing separate tasks, neither one is going to work well (and processes will be worse). It may still be simpler to think in terms of threads, but it'll be the slowest way to do things. You may want to look into coroutines; Greg Ewing has a great demonstration of how to use yield from to use coroutines to build things like schedulers or many-actor simulations.
Next, as I alluded to in your previous question, making threads (or processes) work efficiently with shared state requires holding locks for as short a time as possible.
So, if you have to search a whole queue under a lock, that had better be a constant-time search, not a linear-time search. That's why I suggested using something like an OrderedSet recipe rather than a list, like the one inside the stdlib's Queue.Queue. Then this function:
def is_in_queue(x, q):
with q.mutex:
return x in q.queue
… is only blocking the queue for a tiny fraction of a second—just long enough to look up a hash value in a table, instead of long enough to compare every element in the queue against x.
Finally, I tried to explain about race conditions on your other question, but let me try again.
You need a lock around every complete "transaction" in your code, not just around the individual operations.
For example, if you do this:
with queue locked:
see if x is in the queue
if x was not in the queue:
with queue locked:
add x to the queue
… then it's always possible that x was not in the queue when you checked, but in the time between when you unlocked it and relocked it, someone added it. This is exactly why it's possible for both threads to stop early.
To fix this, you need to put a lock around the whole thing:
with queue locked:
if x is not in the queue:
add x to the queue
Of course this goes directly against what I said before about locking the queue for as short a time as possible. Really, that's what makes multithreading hard in a nutshell. It's easy to write safe code that just locks everything for as long as might conceivably be necessary, but then your code ends up only using a single core, while all the other threads are blocked waiting for the lock. And it's easy to write fast code that just locks everything as briefly as possible, but then it's unsafe and you get garbage values or even crashes all over the place. Figuring out what needs to be a transaction, and how to minimize the work inside those transactions, and how to deal with the multiple locks you'll probably need to make that work without deadlocking them… that's not so easy.

A couple of things that I think can be improved:
Due to the GIL, you might want to use the multiprocessing (rather than threading) module. In general, CPython threading will not cause CPU intensive work to speed up. (Depending on what exactly is the context of your question, it's also possible that multiprocessing won't, but threading almost certainly won't.)
A function like your is_inqueue would likely lead to high contention.
The locked time seems linear in the number of items that need to be traversed:
def is_in_queue(x, q):
with q.mutex:
return x in q.queue
So, instead, you could possibly do the following.
Use multiprocessing with a shared dict:
from multiprocessing import Process, Manager
manager = Manager()
d = manager.dict()
# Fn definitions and such
p1 = Process(target=p1, args=(d,))
p2 = Process(target=p2, args=(d,))
within each function, check for the item like this:
def p1(d):
# Stuff
if 'foo' in d:
return

What kind of problems (if any) would there be combining asyncio with multiprocessing?

As almost everyone is aware when they first look at threading in Python, there is the GIL that makes life miserable for people who actually want to do processing in parallel - or at least give it a chance.
I am currently looking at implementing something like the Reactor pattern. Effectively I want to listen for incoming socket connections on one thread-like, and when someone tries to connect, accept that connection and pass it along to another thread-like for processing.
I'm not (yet) sure what kind of load I might be facing. I know there is currently setup a 2MB cap on incoming messages. Theoretically we could get thousands per second (though I don't know if practically we've seen anything like that). The amount of time spent processing a message isn't terribly important, though obviously quicker would be better.
I was looking into the Reactor pattern, and developed a small example using the multiprocessing library that (at least in testing) seems to work just fine. However, now/soon we'll have the asyncio library available, which would handle the event loop for me.
Is there anything that could bite me by combining asyncio and multiprocessing?

You should be able to safely combine asyncio and multiprocessing without too much trouble, though you shouldn't be using multiprocessing directly. The cardinal sin of asyncio (and any other event-loop based asynchronous framework) is blocking the event loop. If you try to use multiprocessing directly, any time you block to wait for a child process, you're going to block the event loop. Obviously, this is bad.
The simplest way to avoid this is to use BaseEventLoop.run_in_executor to execute a function in a concurrent.futures.ProcessPoolExecutor. ProcessPoolExecutor is a process pool implemented using multiprocessing.Process, but asyncio has built-in support for executing a function in it without blocking the event loop. Here's a simple example:
import time
import asyncio
from concurrent.futures import ProcessPoolExecutor
def blocking_func(x):
time.sleep(x) # Pretend this is expensive calculations
return x * 5
#asyncio.coroutine
def main():
#pool = multiprocessing.Pool()
#out = pool.apply(blocking_func, args=(10,)) # This blocks the event loop.
executor = ProcessPoolExecutor()
out = yield from loop.run_in_executor(executor, blocking_func, 10) # This does not
print(out)
if __name__ == "__main__":
loop = asyncio.get_event_loop()
loop.run_until_complete(main())
For the majority of cases, this is function alone is good enough. If you find yourself needing other constructs from multiprocessing, like Queue, Event, Manager, etc., there is a third-party library called aioprocessing (full disclosure: I wrote it), that provides asyncio-compatible versions of all the multiprocessing data structures. Here's an example demoing that:
import time
import asyncio
import aioprocessing
import multiprocessing
def func(queue, event, lock, items):
with lock:
event.set()
for item in items:
time.sleep(3)
queue.put(item+5)
queue.close()
#asyncio.coroutine
def example(queue, event, lock):
l = [1,2,3,4,5]
p = aioprocessing.AioProcess(target=func, args=(queue, event, lock, l))
p.start()
while True:
result = yield from queue.coro_get()
if result is None:
break
print("Got result {}".format(result))
yield from p.coro_join()
#asyncio.coroutine
def example2(queue, event, lock):
yield from event.coro_wait()
with (yield from lock):
yield from queue.coro_put(78)
yield from queue.coro_put(None) # Shut down the worker
if __name__ == "__main__":
loop = asyncio.get_event_loop()
queue = aioprocessing.AioQueue()
lock = aioprocessing.AioLock()
event = aioprocessing.AioEvent()
tasks = [
asyncio.async(example(queue, event, lock)),
asyncio.async(example2(queue, event, lock)),
]
loop.run_until_complete(asyncio.wait(tasks))
loop.close()

Yes, there are quite a few bits that may (or may not) bite you.
When you run something like asyncio it expects to run on one thread or process. This does not (by itself) work with parallel processing. You somehow have to distribute the work while leaving the IO operations (specifically those on sockets) in a single thread/process.
While your idea to hand off individual connections to a different handler process is nice, it is hard to implement. The first obstacle is that you need a way to pull the connection out of asyncio without closing it. The next obstacle is that you cannot simply send a file descriptor to a different process unless you use platform-specific (probably Linux) code from a C-extension.
Note that the multiprocessing module is known to create a number of threads for communication. Most of the time when you use communication structures (such as Queues), a thread is spawned. Unfortunately those threads are not completely invisible. For instance they can fail to tear down cleanly (when you intend to terminate your program), but depending on their number the resource usage may be noticeable on its own.
If you really intend to handle individual connections in individual processes, I suggest to examine different approaches. For instance you can put a socket into listen mode and then simultaneously accept connections from multiple worker processes in parallel. Once a worker is finished processing a request, it can go accept the next connection, so you still use less resources than forking a process for each connection. Spamassassin and Apache (mpm prefork) can use this worker model for instance. It might end up easier and more robust depending on your use case. Specifically you can make your workers die after serving a configured number of requests and be respawned by a master process thereby eliminating much of the negative effects of memory leaks.

Based on #dano's answer above I wrote this function to replace places where I used to use multiprocess pool + map.
def asyncio_friendly_multiproc_map(fn: Callable, l: list):
"""
This is designed to replace the use of this pattern:
with multiprocessing.Pool(5) as p:
results = p.map(analyze_day, list_of_days)
By letting caller drop in replace:
asyncio_friendly_multiproc_map(analyze_day, list_of_days)
"""
tasks = []
with ProcessPoolExecutor(5) as executor:
for e in l:
tasks.append(asyncio.get_event_loop().run_in_executor(executor, fn, e))
res = asyncio.get_event_loop().run_until_complete(asyncio.gather(*tasks))
return res

See PEP 3156, in particular the section on Thread interaction:
http://www.python.org/dev/peps/pep-3156/#thread-interaction
This documents clearly the new asyncio methods you might use, including run_in_executor(). Note that the Executor is defined in concurrent.futures, I suggest you also have a look there.