dso.rst

Distributed shared objects

Warning

This feature is experimental and as such is unstable. Using it in production is discouraged. Expect errors and API/functionality changes in future releases.

Note

Currently it only works with IBM Cloud Functions

Usage

1. Deploy a custom runtime as follows:

lithops runtime build -f runtime/ibm_cf/Docker.dso id/runtime:tag
lithops runtime create id/runtime:tag

1. Create a DSO server in the Cloud following the instructions available here: https://github.com/crucial-project/dso/tree/2.0
2. Run a script using a command of the form "DSO=IP:11222" python3 my_script.py, where DSO is the address of a running DSO deployment.

Examples

from dso.client import Client
from jpype import *
from jpype import java
import lithops
import os

dso = os.environ.get('DSO')


def my_map_function(id, x):
    client = Client(dso)
    c = client.getAtomicCounter("cnt")
    c.increment(x)
    b = client.getCyclicBarrier("b", len(iterdata))
    b.waiting()
    return c.tally()


if __name__ == "__main__":
    f = Client(dso)
    c = f.getAtomicCounter("cnt")
    c.reset()
    iterdata = [1, 2, 3, 4]
    fexec = lithops.FunctionExecutor(runtime='0track/lithops-dso:1.1')
    fexec.map(my_map_function, iterdata)
    print(fexec.get_result())
    print(c.tally())
    fexec.clean()

from dso.client import Client
from jpype import *
from jpype import java
import lithops
import os

dso = os.environ.get('DSO')


def my_function(x):
    client = Client(dso)
    d = client.getAtomicCounter("cnt")
    return d.increment(x)


if __name__ == '__main__':
    fexec = lithops.FunctionExecutor(runtime='0track/lithops-dso:1.1')
    fexec.call_async(my_function, 3)
    client = Client(dso)
    c = client.getAtomicCounter("cnt")
    print("counter: " + str(c.tally()))
    print(fexec.get_result())
    print("counter: " + str(c.tally()))