How to deploy scikit-hts at scale¶

The high volume of time series data can demand an analysis at scale. So, time series need to be processed on a group of computational units instead of a singular machine.

Accordingly, it may be necessary to distribute the extraction of time series features to a cluster. Indeed, it is possible to extract features with hts in a distributed fashion. This page will explain how to setup a distributed hts.

The distributor class¶

To distribute the calculation of features, we use a certain object, the Distributor class (contained in the hts.utilities.distribution module).

Essentially, a Distributor organizes the application of feature calculators to data chunks. It maps the feature calculators to the data chunks and then reduces them, meaning that it combines the results of the individual mapping into one object, the feature matrix.

So, Distributor will, in the following order,

calculates an optimal chunk_size, based on the characteristics of the time series data at hand (by calculate_best_chunk_size())

split the time series data into chunks (by partition())

distribute the applying of the feature calculators to the data chunks (by distribute())

combine the results into the feature matrix (by map_reduce())

close all connections, shutdown all resources and clean everything (by close())

So, how can you use such a Distributor to extract features with hts? You will have to pass it into as the distributor argument to the extract_features() method.

The following example shows how to define the MultiprocessingDistributor, which will distribute the calculations to a local pool of threads:

from hts import HTSRegressor
from hts.utilities.load_data import load_mobility_data
from hts.utilities.distribution import MultiprocessingDistributor

df = load_mobility_data()

# Define hierarchy
hier = {
    'total': ['CH', 'SLU', 'BT', 'OTHER'],
    'CH': ['CH-07', 'CH-02', 'CH-08', 'CH-05', 'CH-01'],
    'SLU': ['SLU-15', 'SLU-01', 'SLU-19', 'SLU-07', 'SLU-02'],
    'BT': ['BT-01', 'BT-03'],
    'OTHER': ['WF-01', 'CBD-13']
}

distributor = MultiprocessingDistributor(n_workers=4,
                                         disable_progressbar=False,
                                         progressbar_title="Feature Extraction")
hts.fit(df=df, nodes=hier, n_jobs=4, distributor=distributor)

This example actually corresponds to the existing multiprocessing API, where you just specify the number of jobs, without the need to construct the Distributor:

from hts import HTSRegressor
from hts.utilities.load_data import load_mobility_data

df = load_mobility_data()

# Define hierarchy
hier = {
    'total': ['CH', 'SLU', 'BT', 'OTHER'],
    'CH': ['CH-07', 'CH-02', 'CH-08', 'CH-05', 'CH-01'],
    'SLU': ['SLU-15', 'SLU-01', 'SLU-19', 'SLU-07', 'SLU-02'],
    'BT': ['BT-01', 'BT-03'],
    'OTHER': ['WF-01', 'CBD-13']
}

hts.fit(df=df, nodes=hier, n_jobs=4)

Using dask to distribute the calculations¶

We provide distributor for the dask framework, where “Dask is a flexible parallel computing library for analytic computing.”

Dask is a great framework to distribute analytic calculations to a cluster. It scales up and down, meaning that you can even use it on a singular machine. The only thing that you will need to run hts on a Dask cluster is the ip address and port number of the dask-scheduler.

Lets say that your dask scheduler is running at 192.168.0.1:8786, then we can easily construct a ClusterDaskDistributor that connects to the scheduler and distributes the time series data and the calculation to a cluster:

from hts import HTSRegressor
from hts.utilities.load_data import load_mobility_data
from hts.utilities.distribution import ClusterDaskDistributor


df = load_mobility_data()

# Define hierarchy
hier = {
    'total': ['CH', 'SLU', 'BT', 'OTHER'],
    'CH': ['CH-07', 'CH-02', 'CH-08', 'CH-05', 'CH-01'],
    'SLU': ['SLU-15', 'SLU-01', 'SLU-19', 'SLU-07', 'SLU-02'],
    'BT': ['BT-01', 'BT-03'],
    'OTHER': ['WF-01', 'CBD-13']
}

distributor = ClusterDaskDistributor(address="192.168.0.1:8786")
hts.fit(df=df, nodes=hier)
...

# Prediction also runs in a distributed fashion
preds = hts.predict(steps_ahead=10)

Compared to the MultiprocessingDistributor example from above, we only had to change one line to switch from one machine to a whole cluster. It is as easy as that. By changing the Distributor you can easily deploy your application to run to a cluster instead of your workstation.

You can also use a local DaskCluster on your local machine to emulate a Dask network. The following example shows how to setup a LocalDaskDistributor on a local cluster of 3 workers:

from hts import HTSRegressor
from hts.utilities.load_data import load_mobility_data
from hts.utilities.distribution import LocalDaskDistributor


df = load_mobility_data()

# Define hierarchy
hier = {
    'total': ['CH', 'SLU', 'BT', 'OTHER'],
    'CH': ['CH-07', 'CH-02', 'CH-08', 'CH-05', 'CH-01'],
    'SLU': ['SLU-15', 'SLU-01', 'SLU-19', 'SLU-07', 'SLU-02'],
    'BT': ['BT-01', 'BT-03'],
    'OTHER': ['WF-01', 'CBD-13']
}

distributor = LocalDaskDistributor(n_workers=3)
hts.fit(df=df, nodes=hier)
...

# Prediction also runs in a distributed fashion
preds = hts.predict(steps_ahead=10)

Writing your own distributor¶

If you want to user another framework than Dask, you will have to write your own Distributor. To construct your custom Distributor, you will have to define an object that inherits from the abstract base class hts.utilities.distribution.DistributorBaseClass. The hts.utilities.distribution module contains more information about what you will need to implement.

Acknowledgement¶

This documentation, as well as the underlying implementation, exists only thanks to the folks at blue-yonder. The This page was pretty much copy and pasted from their tsfresh package. Many thanks for their excellent package.