Fastmap is a drop-in replacement for `map` that parallelizes your code on the cloud.
Project description
Note: Fastmap is currently in beta.
Fastmap is a drop-in replacement for map
that makes your Python code in parallel on the cloud. Fastmap is appropriate to use when map
is too slow but setting up custom infrastructure would be overkill.
- 🚀 Speed up parallel tasks. Fastmap automatically parallelizes your code and distributes work locally and on the cloud.
- 🐣 Trivial to use. Add
global_init
to the top of your file, and replace every instance ofmap
withfastmap
. There is no code to upload and the SDK consists of only 3 functions. - 🐣 Free and open. Fastmap is open source, transparent, and simple. Don't get locked into proprietary frameworks or, for that matter, your own infrastructure.
- 🚀 Deploy in minutes. With a Google Cloud Platform account, you can setup and deploy your fastmap cloud service with one command.
Docs
For complete documentation, go to https://fastmap.io/docs,
SDK installation
pip3 install -U fastmap
Conceptual local example
This maps your code locally across multiple CPUs. For a LOCAL exec policy, no extra server setup is required.
import csv
from my_project import big_function
import fastmap
config = fastmap.init(exec_policy="LOCAL")
with open('lots_of_data.csv') as fh:
long_list = list(csv.reader(fh))
results = list(config.fastmap(big_function, long_list))
Conceptual cloud example
To setup a server to test with, see https://github.com/fastmap-io/fastmap-server. This can be deployed either locally or to GCP. After running the single-command deploy script, you will have your CLOUD_URL and SECRET_TOKEN.
Important: Protect your secret token like a password and never commit it to version control
import csv
from config import CLOUD_URL, SECRET_TOKEN
from my_project import big_function
import fastmap
config = fastmap.init(
cloud_url=CLOUD_URL,
secret=SECRET_TOKEN)
with open('lots_of_data.csv') as fh:
long_list = list(csv.reader(fh))
results = list(config.fastmap(big_function, long_list))
Runnable example
See fastmap_example_test.py on the open source cloud service repo. This will estimate pi using the Monte Carlo method.
When should you use fastmap?
As a rule-of-thumb, fastmap will speed up any call to map that would have otherwise taken more than one second. This is possible because, under the default ADAPTIVE execution policy, fastmap algorithmically distributes work locally and across the cloud.
If you are planning to use the 'CLOUD' exec_policy, which prevents local processing, fastmap is appropriate when your function is either a scraper or is computationally-heavy. This is because transferring data to the cloud for processing always takes a non-zero amount of time. The trade-off depends on your network speeds and distance to your fastmap server cluster.
If in doubt, try running fastmap with a small test dataset. Fastmap attempts to be transparent and will inform you when using it has made your code slower.
Questions
Fastmap.io is a new project and I would love to hear your feedback. You can contact me directly at scott@fastmap.io.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.