Python API via FSSpec

This feature is experimental.

Alluxio Python SDK (alluxiofs) is based on FSSpec, which allows applications to seamlessly interact with various storage backends using a unified Python filesystem interface.It leverages a high-performance distributed caching layer, the Alluxio cluster, to significantly enhance data access speeds and reduce latency.This is particularly beneficial for data-intensive applications and workflows, especially AI training workloads, where large datasets need to be accessed quickly and repeatedly.

Prerequisites

• A running Alluxio cluster

• Python's version >= 3.8

Installation

Install Dependencies

Install alluxiofs

pip install alluxiofs

(Optional) Install UFS's client base on FSSpec(such as s3, oss)

Example: If your UFS is S3, and you want to have the ability to fallback to UFS after Alluxio fails, you need to download s3fs

pip install s3fs

Environment Setup

1. Start an Alluxio cluster with at least 1 coordinator and 1 worker

• If you want to start cluster in K8s, please refer to Alluxio Official Documentation).

• If you want to start cluster in Bare Metal, you can use ./bin/alluxio process start cli.

1. Interacting with alluxio using alluxiofs

For details，you can go to Create alluxiofs's Instance and alluxiofs's Basic File Operations.

Create alluxiofs's Instance

This simple example creates a filesystem handler that connects to an Alluxio cluster, using "localhost" as load balance domain, using S3 as the under storage.

import fsspec
from alluxiofs import AlluxioFileSystem, setup_logger

# Register Alluxio to fsspec
fsspec.register_implementation("alluxiofs", AlluxioFileSystem, clobber=True)

# Create Alluxio filesystem
alluxio_options = {
    "alluxio.worker.page.store.page.size": "4MB",
}
alluxio_fs = fsspec.filesystem(
    "alluxiofs",
    load_balance_domain="localhost", #or worker_hosts="127.0.0.1, 127.0.0.2"
    target_protocol="s3",
    options=alluxio_options
    logger=setup_logger("./", level=logging.DEBUG),
)

See advanced init arguments to set advanced arguments for Alluxio cluster and/or storage system connections.

alluxiofs's Basic File Operations

Below are common operations supported by alluxio_fs for interacting with files and directories:

1. ls

ls(self, path: str, detail: bool=False) -> list[dict]

Lists the contents of a directory.

Parameters

• path (str): The path of the directory to list.

• detail (bool, optional): If True, returns detailed information about each entry. Defaults to False.

Returns

• list[dict]: A list of dicts with JSON-like structure.

Example

contents = alluxio_fs.ls("/path/to/directory", False)
print(contents)

2. info

info(self, path: str) -> dict

Retrieves information about a file or directory.

Parameters

• path (str): The path of the file or directory.

Returns

• dict: JSON-like structure with file or directory info.

Example

file_info = alluxio_fs.info("/path/to/file")
print(file_info)

3. isdir

isdir(self, path: str) -> bool

Checks if a path is a directory.

Parameters

• path (str): The path to check.

Returns

• bool: Whether the path is a directory.

Example

is_directory = alluxio_fs.isdir("/path/to/directory")
print(is_directory)

4. open

open(self, path: str, mode: str="rb", block_size: int=None, autocommit: bool=True, cache_options: dict=None, **kwargs) -> AlluxioFile

Opens a file for reading or writing.

Parameters

• path (str): The path of the file to open.

• mode (str, optional): The mode in which to open the file. Defaults to "rb".

• block_size (int, optional): The block size for reading.

• autocommit (bool, optional): If True, commits changes automatically.

• cache_options (dict, optional): Cache options.

• **kwargs: Additional keyword arguments.

Returns

• AlluxioFile: Object supporting read(), write(), etc.

Example

with alluxio_fs.open("/path/to/file", mode="rb") as f:
    data = f.read()
    print(data)

with alluxio_fs.open("/path/to/file", mode="wb") as f:
    f.write(data)

5. mkdir

mkdir(self, path: str) -> bool

Creates a directory.

Parameters

• path (str): The path of the new directory.

Returns

• bool: Whether the directory was created successfully.

Example

alluxio_fs.mkdir("/path/to/new_directory")

6. rm

rm(self, path: str, recursive: bool=False, recursiveAlias: bool=False, removeAlluxioOnly: bool=False, deleteMountPoint: bool=False, syncParentNextTime: bool=False, removeUncheckedOptionChar: bool=False) -> bool

Removes a file or directory.

Parameters: Multiple optional flags for control.

Returns

• bool: Whether the operation succeeded.

Example

alluxio_fs.rm("/path/to/file_or_directory", recursive=True)

7. touch

touch(self, path: str) -> bool

Creates an empty file.

Parameters

• path (str): File path.

Returns

• bool: Whether the file was created.

Example

alluxio_fs.touch("/path/to/new_file")

8. head

head(self, path: str, num_of_bytes: int=1024) -> bytes

Reads the first few bytes of a file.

Parameters

• path (str): File path.

• num_of_bytes (int): Number of bytes.

Returns

• bytes: The read bytes.

Example

data = alluxio_fs.head("/path/to/file", num_of_bytes=1024)
print(data)

9. tail

tail(self, path: str, num_of_bytes: int=1024) -> bytes

Reads the last few bytes of a file.

Parameters

• path (str): File path.

• num_of_bytes (int): Number of bytes.

Returns

• bytes: The read bytes.

Example

data = alluxio_fs.tail("/path/to/file", num_of_bytes=1024)
print(data)

10. mv

mv(self, path1: str, path2: str) -> bool

Moves or renames a file or directory.

Parameters

• path1 (str): Source path.

• path2 (str): Destination path.

Returns

• bool: Whether the move/rename succeeded.

Example

alluxio_fs.mv("/path/to/source", "/path/to/destination")

11. cp

cp(self, path1: str, path2: str, recursive: bool=False, recursiveAlias: bool=False, force: bool=False, thread: int=None, bufferSize: int=None, preserve: bool=None) -> bool

Copies a file or directory.

Parameters: Various options for recursion, force, threading, etc.

Returns

• bool: Whether the copy succeeded.

Example

alluxio_fs.cp("/path/to/source", "/path/to/destination", recursive=True)

12. read_bytes

read_bytes(self, path: str) -> bytes

Reads an entire file (for little files <10MB).

Parameters

• path (str): File path.

Returns

• bytes: The read content.

Example

data = alluxio_fs.read_bytes("/path/to/file")
print(data)

13. write_bytes

write_bytes(self, path: str, value: bytes) -> bool

Writes byte data to a file in Alluxio.

Parameters

• path (str): Path to write the data to.

• value (bytes): The byte content.

Returns

• bool: Whether the write succeeded.

Example

alluxio_fs.write_bytes("/path/to/file", b"data to write")

14. upload

upload(self, lpath: str, rpath: str) -> bool

Upload a large file from the local OS file system to Alluxio (and UFS, depending on the WriteType).

Parameters

• lpath (str): The source path of the file in the local OS.

• rpath (str): The destination path of the file in Alluxio/UFS.

Returns

• bool: Whether the upload task completed successfully.

Example

# upload a large file to Alluxio
alluxio_fs.upload("/path/in/local", "/path/in/UFS")

15. download

download(self, lpath: str, rpath: str) -> bool

Downloads a large file from Alluxio to the local OS file system.

Parameters

• lpath (str): The destination path in the local file system.

• rpath (str): The source path in Alluxio.

Returns

• bool: Whether the download task completed successfully.

Example

# download a large file from Alluxio
alluxio_fs.download("/local/path", "/alluxio/path")

More Python filesystem operations examples can be found here.

Integration with Other Frameworks

Example: Ray

Ray is a fast and simple framework for building and running distributed applications. PyTorch, TensorFlow, and XGBoost trainers, running on top of Ray, can leverage Ray's advanced functionalities like creating heterogeneous clusters consisting of CPU machines for data loading and preprocessing, and GPU machines for training. Data loading, preprocessing, and training can be parallelized using Ray Data.

Trainers like PyTorch will read the same dataset again and again for each epoch. Getting a large dataset for PyTorch in each epoch becomes the training bottleneck. By leveraging Alluxio's high-performance distributed caching, trainers on Ray can reduce total training time, improve GPU utilization rate, and speed up the end-to-end model lifecycle.

Prerequisites: Ray version >= 2.8.2

# Pass the initialized Alluxio filesystem to Ray and read the dataset
ds = ray.data.read_parquet("s3://bucket/path/to/dataset/file1.parquet", filesystem=alluxio_fs)

# Get a count of the number of records in the single file
ds.count()

# Display the schema derived from the file header record
ds.schema()

# Display the header record
ds.take(1)

# Display the first data record
ds.take(2)

# Read multiple files:
ds2 = ray.data.read_parquet("s3://bucket/path/to/dataset/", filesystem=alluxio_fs)

# Get a count of the number of records in the files
ds2.count()

Example: PyArrow

PyArrow allows applications and data to seamlessly connect with each other by providing a high-performance, in-memory columnar storage format. It enables efficient data interchange between different data processing systems. By delegating its storage interface to fsspec, PyArrow can access various storage backends through a unified interface. By using alluxiofs, PyArrow can leverage Alluxio's distributed caching capabilities to enhance data access speeds and reduce latency.

Example 1:

# Pass the initialized Alluxio filesystem to Pyarrow and read the data set from the example csv file
import pyarrow.dataset as ds
dataset = ds.dataset("s3://bucket/path/to/dataset/file1.parquet", filesystem=alluxio_fs)

# Get a count of the number of records in the file
dataset.count_rows()

# Display the schema derived from the file header record
dataset.schema

# Display the first record
dataset.take(0)

Example 2:

from pyarrow.fs import PyFileSystem, FSSpecHandler

# Create a python-based PyArrow filesystem using FsspecHandler
py_fs = PyFileSystem(FSSpecHandler(alluxio_fs))

# Read the data by using the Pyarrow filesystem interface
with py_fs.open_input_file("s3://bucket/path/to/dataset/file1.parquet") as f:
    f.read()

Advanced Initialization Parameters

Arguments to Connect to Alluxio Cluster

• options (dict, optional): A dictionary of Alluxio configuration options where keys are property names and values are property values. These options configure the Alluxio client behavior.

Example: Configure Alluxio fsspec. Note that the following options must be the same between alluxiofs and alluxio cluster

• alluxio.worker.page.store.page.size (default 1MB): Size of each page in worker paged block store. Recommend to set to 20MB for large parquet files.

• alluxio.user.worker.selection.policy.consistent.hash.virtual.node.count.per.worker (default 2000): This is the number of virtual nodes for one worker in the consistent hashing algorithm. In a consistent hashing algorithm, on membership changes, some virtual nodes are re-distributed instead of rebuilding the whole hash table. This guarantees the hash table is changed only in a minimal. In order to achieve that, the number of virtual nodes should be X times the physical nodes in the cluster, where X is a balance between redistribution granularity and size. Recommend to set to 5.

alluxio_options = {
    "alluxio.user.worker.selection.policy.consistent.hash.virtual.node.count.per.worker": "2000",
    "alluxio.worker.page.store.page.size": "4MB",
}

(Optional) Init alluxio_client for distributed load operations:

alluxio_client = AlluxioClient(
    load_balance_domain="localhost", #or worker_hosts="127.0.0.1, 127.0.0.2" 
    options=alluxio_options
)

Init alluxio_fs for fsspec filesystem operations:

alluxio_fs = fsspec.filesystem(
    "alluxiofs", 
    load_balance_domain="localhost", #or worker_hosts="127.0.0.1, 127.0.0.2"
    target_protocol="s3", 
    options=alluxio_options
)

config the logger of alluxio_fs

import logging
import fsspec
from alluxiofs import AlluxioFileSystem, setup_logger

alluxio_fs = fsspec.filesystem(
    "alluxiofs", 
    load_balance_domain="localhost", #or worker_hosts="127.0.0.1, 127.0.0.2"
    logger=setup_logger("./", level=logging.DEBUG)
)

Arguments for storage backend

Arguments:

• target_protocol (str, optional): Specifies the under storage protocol to create the under storage file system object. Common examples include s3 for Amazon S3, hdfs for Hadoop Distributed File System, and others.

• target_options (dict, optional): Provides a set of configuration options relevant to the target_protocol. These options might include credentials, endpoint URLs, and other protocol-specific settings required to successfully interact with the under storage system.

• fs (object, optional): Directly supplies an instance of a file system object for accessing the underlying storage of Alluxio

• logger(object， optional): config the the path to store log files and level of logger, the path is the current path default, and the level is logging.INFO default.

Example: connect to S3

To connect to S3, you can follow these steps:

1. Review S3 fsspec documentation: Refer to the s3fs documentation to find out the supported arguments for connecting to S3. Typical arguments include:

• anon bool (False): Whether to use anonymous connection (public buckets only). If False, uses the key/secret given, or boto's credential resolver; client_kwargs, environment, variables, config files, EC2 IAM server, in that order

• endpoint_url string (None): Use this endpoint_url, if specified. Needed for connecting to non-AWS S3 buckets. Takes precedence over endpoint_url in client_kwargs.

• key string (None): If not anonymous, use this access key ID, if specified. Takes precedence over aws_access_key_id in client_kwargs.

• secret string (None): If not anonymous, use this secret access key, if specified. Takes precedence over aws_secret_access_key in client_kwargs.

• token string (None): If not anonymous, use this security token, if specified

1. Pass the supported arguments as target_options to Alluxio: You can then use these arguments to create an Alluxio file system object using fsspec.

Here's how to create an Alluxio file system object connected to S3 using fsspec:

import fsspec

# Example configuration for connecting to S3
s3_options = {
    "key": "your-aws-access-key",
    "secret": "your-aws-secret-key",
    "endpoint_url": "https://s3.your-endpoint.com"
}

# Creating the Alluxio file system object
alluxio_fs = fsspec.filesystem(
    "alluxiofs",
    load_balance_domain="localhost", #or worker_hosts="127.0.0.1, 127.0.0.2"
    target_protocol="s3",
    target_options=s3_options
)

# Now you can use alluxio_fs to interact with the Alluxio file system backed by S3

In this example:

• Replace your-aws-access-key and your-aws-secret-key with your actual AWS credentials.

• Replace https://s3.your-endpoint.com with the appropriate endpoint URL for your S3-compatible service if needed.

By following these steps, you can effectively connect to Alluxio with an S3 backend using fsspec.

Monitoring Metrics

Monitoring System Setup

1. Prometheus Installation and Configuration Edit the prometheus.yml configuration file and then start Prometheus:

   nohup ./prometheus --web.enable-admin-api --config.file=./prometheus.yml >./prometheus.log 2>&1 &

2. Grafana Installation and Configuration Start Grafana:

   nohup ./bin/grafana-server --homepath . web >./grafana.log 2>&1 &

For more detailed instructions on setting up the monitoring system, refer to the Alluxio Official Documentation.

Explanation of Monitoring Metrics

Metric Name	Description	Labels	Unit	Implementation Code
alluxio_http_server_call_latency_ms	Histogram of HTTP service call latency (Bucket boundaries: [10, 40, 160, 640] ms)	method, success	Milliseconds (ms)	HistogramWrapper
alluxio_http_server_result_total	Total count of HTTP service results	method, state	Count	CounterWrapper
alluxio_http_server_call_latency_ms_sum	Total latency of HTTP service calls	method, success	Milliseconds (ms)	HistogramWrapper
alluxio_http_server_call_latency_ms_count	Count of HTTP service calls	method, success	Count	HistogramWrapper