您可以通过探索多线程来获得更好的上传性能。这里有一些代码可以做到这一点:
from azure.storage.blob import BlobClient
from threading import Thread
import os
# Uploads a single blob. May be invoked in thread.
def upload_blob(container, file, index=0, result=None):
if result is None:
result = [None]
try:
# extract blob name from file path
blob_name = ''.join(os.path.splitext(os.path.basename(file)))
blob = BlobClient.from_connection_string(
conn_str='CONNECTION STRING',
container_name=container,
blob_name=blob_name
)
with open(file, "rb") as data:
blob.upload_blob(data, overwrite=True)
print(f'Upload succeeded: {blob_name}')
result[index] = True # example of returning result
except Exception as e:
print(e) # do something useful here
result[index] = False # example of returning result
# container: string of container name. This example assumes the container exists.
# files: list of file paths.
def upload_wrapper(container, files):
# here, you can define a better threading/batching strategy than what is written
# this code just creates a new thread for each file to be uploaded
parallel_runs = len(files)
threads = [None] * parallel_runs
results = [None] * parallel_runs
for i in range(parallel_runs):
t = Thread(target=upload_blob, args=(container, files[i], i, results))
threads[i] = t
threads[i].start()
for i in range(parallel_runs): # wait for all threads to finish
threads[i].join()
# do something with results here
可能有更好的分块策略 - 这只是一个示例,说明在某些情况下,您可以通过使用线程来实现更高的 Blob 上传性能。
以下是顺序循环方法与上述线程方法(482 个图像文件,总共 26 MB)之间的一些基准:
我还应该补充一点,您可以考虑通过 Python 调用 azcopy,因为此工具可能更适合您的特定需求。