Search code examples
pythonsshsftpparamikonetcdf

Reading file opened with Python Paramiko SFTPClient.open method is slow

I am trying to remote read a netcdf file.
I used Paramiko package to read my file, like this:

import paramiko
from netCDF4 import Dataset

client = paramiko.SSHClient()
client.set_missing_host_key_policy(paramiko.AutoAddPolicy())
client.connect(hostname=’hostname’, username=’usrname’, password=’mypassword’)

sftp_client = client.open_sftp()
ncfile = sftp_client.open('mynetCDFfile')
b_ncfile = ncfile.read()    # ****

nc = Dataset('test.nc', memory=b_ncfile)

But the run speed of ncfile.read() is VERY SLOW.

So my question is: Is there any alternative way to read a netcdf file remotely, or is there any approach to speed up paramiko.sftp_file.SFTPFile.read()?

Solution

  • Calling SFTPFile.prefetch should increase the read speed:

    ncfile = sftp_client.open('mynetCDFfile')
ncfile.prefetch()
b_ncfile = ncfile.read()

    Another option is enabling read buffering, using bufsize parameter of SFTPClient.open:

    ncfile = sftp_client.open('mynetCDFfile', bufsize=32768)
b_ncfile = ncfile.read()

    (32768 is a value of SFTPFile.MAX_REQUEST_SIZE)

    Similarly for writes/uploads:
    Writing to a file on SFTP server opened using Paramiko/pysftp "open" method is slow.

    Yet another option is to explicitly specify the amount of data to read (it makes BufferedFile.read take a more efficient code path):

    ncfile = sftp_client.open('mynetCDFfile')
b_ncfile = ncfile.read(ncfile.stat().st_size)

    If none of that works, you can download the whole file to memory instead:
    Use pdfplumber and Paramiko to read a PDF file from an SFTP server

    Obligatory warning: Do not use AutoAddPolicy this way – You are losing a protection against MITM attacks by doing so. For a correct solution, see Paramiko "Unknown Server".