Web scraping at scale often requires big pools of proxies to avoid rate limiting, throttling or even blocking. As proxies are expensive resources correct distribution can make a huge difference in the overall resource usage.
In this article, we'll take at what is proxy rotation and how it helps to avoid blocking as well as some common rotation patterns, tips and tricks.
When web scraping with a pool of proxies our connection success chances can be influenced by connection patterns. In other words, if proxy A connects to 50 pages in 5 seconds it's like to get throttled or blocked. However, if we have proxies A, B, C and have them take turns it will avoid pattern-based rate limiting and blocking.
So, what are the best ways to rotate our proxies?
Proxy Distribution and Order
The way we distribute our proxy rotation can make a huge difference. For example, a common approach is to simply pluck a random proxy from the pool for each request however it's not a very smart or efficient approach.
The first thing we should do is rotate proxies by subnet or proxy origin. For example, if we have a proxy pool of:
import random
proxies = [
"xx.xx.123.1",
"xx.xx.123.2",
"xx.xx.123.3",
"xx.xx.124.1",
"xx.xx.124.2",
"xx.xx.125.1",
"xx.xx.125.2",
]
def get_proxy():
return random.choice(proxies)
print("got 10 proxies:")
for i in range(10):
print(get_proxy())
If we were to pluck a random proxy for our web scraping request there's a likely chance that proxies of the same subnet would end up in a row.
The subnet is the 3rd number of a IPv4 address
As many throttling/blocking services consider subnet in their logic our proxy rotator would be in a major disadvantage.
Instead, it's best to ensure that we randomize by subnet:
import random
proxies = [
"xx.xx.123.1",
"xx.xx.123.2",
"xx.xx.123.3",
"xx.xx.124.1",
"xx.xx.124.2",
"xx.xx.125.1",
"xx.xx.125.2",
]
last_subnet = ""
def get_proxy():
while True:
ip = random.choice(proxies)
ip_subnet = ip.split('.')[2]
if ip_subnet != last_subnet:
last_subnet = ip_subnet
return ip
print("got 10 proxies:")
for i in range(10):
print(get_proxy())
Now, our proxy selector will never return two proxies from the same subnet in a row.
The subnet is not the only axis that should weigh on randomization.
Proxies also come with a lot of metadata such as IP address ASN (Autonomous System Number which essentially means "proxy IP owner ID"), location etc. Based on the web scraping target it's a good idea to rotate proxies based on all of these features.
Once we have our proxies rolling eventually they will start being blocked. Some will be more successful than others and some blocked ones will at some point recover.
We can further extend our rotator to track proxy performance. For example, we can mark dead ones to prevent from being picked so they have some time to recover:
import random
from datetime import datetime
proxies = [
"xx.xx.123.1",
"xx.xx.123.2",
"xx.xx.123.3",
"xx.xx.124.1",
"xx.xx.124.2",
"xx.xx.125.1",
"xx.xx.125.2",
]
last_subnet = ""
dead_proxies = {}
def get_proxy():
while True:
ip = random.choice(proxies)
ip_subnet = ip.split('.')[2]
if ip_subnet == last_subnet:
continue
if ip in dead_proxies:
if dead_proxies[ip] - datetime.utcnow() > timedelta(seconds=30):
#proxy has not recovered yet - skip
continue
else:
# proxy has recovered - set it free!
del dead_proxies[ip]
last_subnet = ip_subnet
return ip
def scrape(url, retries=0):
proxy = get_proxy()
response = requests.get(url, proxies=proxy)
# mark dead proxies and retry
if response.status_code == 200:
return response
else:
dead_proxies[proxy] = datetime.utcnow()
retries += 1
if retries > 3:
raise RetriesExceeded()
scrape(url, retries=retries)
Above, our simple proxy rotator distributes proxies randomly by subnet and keeps track of dead proxies.
Weighted Random Proxy Example Rotator
Let's put together what we've learned and build a random rotator. For this we'll be using python weighted random choice function random.choices.
random.choices selects a random element but the best feature is that we can assign custom weights to the available choices. This allows us to prioritize some proxies over the others. For example:
from collections import Counter
import random
proxies = [
("xx.xx.123.1", 10),
("xx.xx.123.2", 1),
("xx.xx.123.3", 1),
("xx.xx.124.1", 10),
("xx.xx.124.2", 1),
("xx.xx.125.1", 10),
("xx.xx.125.2", 1),
]
counter = Counter()
for i in range(1000):
choice = random.choices([p[0] for p in proxies], [p[1] for p in proxies], k=1)
counter[choice[0]] += 1
for proxy, used_count in counter.most_common():
print(f"{proxy} was used {used_count} times")
In the above example, we weighted all proxies that end in .1 ten times as heavy as others - which means they get picked 10 times more often:
xx.xx.125.1 was used 298 times
xx.xx.124.1 was used 292 times
xx.xx.123.1 was used 283 times
xx.xx.125.2 was used 38 times
xx.xx.124.2 was used 34 times
xx.xx.123.3 was used 30 times
xx.xx.123.2 was used 25 times
Weighted randomization gives us immense creativity in designing our proxy rotator!
For example, we can:
penalize recently used proxies
penalize failing proxies
promote healthy or fast proxies
promote one proxy type over the other (e.g. residential proxies over datacenter)
This approach allows us to create proxy rotators that are unpredictable but smart. Let's take a look at an example implementing this logic:
import random
from time import time
from typing import List, Literal
class Proxy:
"""container for a proxy"""
def __init__(self, ip, type_="datacenter") -> None:
self.ip: str = ip
self.type: Literal["datacenter", "residential"] = type_
_, _, self.subnet, self.host = ip.split(":")[0].split('.')
self.status: Literal["alive", "unchecked", "dead"] = "unchecked"
self.last_used: int = None
def __repr__(self) -> str:
return self.ip
def __str__(self) -> str:
return self.ip
class Rotator:
"""weighted random proxy rotator"""
def __init__(self, proxies: List[Proxy]):
self.proxies = proxies
self._last_subnet = None
def weigh_proxy(self, proxy: Proxy):
weight = 1_000
if proxy.subnet == self._last_subnet:
weight -= 500
if proxy.status == "dead":
weight -= 500
if proxy.status == "unchecked":
weight += 250
if proxy.type == "residential":
weight += 250
if proxy.last_used:
_seconds_since_last_use = time() - proxy.last_used
weight += _seconds_since_last_use
return weight
def get(self):
proxy_weights = [self.weigh_proxy(p) for p in self.proxies]
proxy = random.choices(
self.proxies,
weights=proxy_weights,
k=1,
)[0]
proxy.last_used = time()
self.last_subnet = proxy.subnet
return proxy
Example Run Code & Output
We can mock-run our Rotator to see proxy distribution:
from collections import Counter
if __name__ == "__main__":
proxies = [
# these will be used more often
Proxy("xx.xx.121.1", "residential"),
Proxy("xx.xx.121.2", "residential"),
Proxy("xx.xx.121.3", "residential"),
# these will be used less often
Proxy("xx.xx.122.1"),
Proxy("xx.xx.122.2"),
Proxy("xx.xx.123.1"),
Proxy("xx.xx.123.2"),
]
rotator = Rotator(proxies)
# let's mock some runs:
_used = Counter()
_failed = Counter()
def mock_scrape():
proxy = rotator.get()
_used[proxy.ip] += 1
if proxy.host == "1": # simulate proxies with .1 being significantly worse
_fail_rate = 60
else:
_fail_rate = 20
if random.randint(0, 100) < _fail_rate: # simulate some failure
_failed[proxy.ip] += 1
proxy.status = "dead"
mock_scrape()
else:
proxy.status = "alive"
return
for i in range(10_000):
mock_scrape()
for proxy, count in _used.most_common():
print(f"{proxy} was used {count:>5} times")
print(f" failed {_failed[proxy]:>5} times")
Now, when we run our script we can see that our rotator picks residential proxies more often and proxies ending with ".1" less often just by using weighted randomization:
xx.xx.121.2 was used 2629 times
failed 522 times
xx.xx.121.3 was used 2603 times
failed 508 times
xx.xx.123.2 was used 2321 times
failed 471 times
xx.xx.122.2 was used 2302 times
failed 433 times
xx.xx.121.1 was used 1941 times
failed 1187 times
xx.xx.122.1 was used 1629 times
failed 937 times
xx.xx.123.1 was used 1572 times
failed 939 times
In this example, we can see how we can use simple probability adjustment to intelligently rotate proxies with very little oversight or tracking code.
Proxies with ScrapFly
ScrapFly web scraping API has over 190M of proxies from 92+ countries at its disposal and rotates IPs smartly for us.
from scrapfly import ScrapflyClient, ScrapeConfig
client = ScrapflyClient(key="YOUR SCRAPFLY KEY")
client.scrape(ScrapeConfig(
url="https://httpbin.org/get",
# we can enable Anti Scraping Protection bypass:
asp=True,
))
The API will smartly select a proxy and a connection fingerprint that will be let through by the anti-scraping service.
Proxy Rotation Summary
In this article, we've taken a look at common proxy rotation strategies and how they can help us to avoid blocking when web scraping.
We covered a few popular rotation patterns like score tracking based on proxy details and performance. Finally, we made an example proxy rotator hat uses random weighted proxy selection to smartly select proxies for our web scraping connections.
