Performance Comparison of Linux File Transfer Tools (scp, ftp, sftp, rsync, tftp, nc, Python SimpleHTTPServer)

The author, a DBA from the iKOS delivery service team, wanted to identify the most efficient data‑transfer tool for Linux servers. After a brief search, six tools were selected: ftp, sftp, scp, rsync, tftp, nc , plus a Python SimpleHTTPServer solution.

Test Environment and Preparation

Two Linux machines were configured with SSH trust. Two test directories were created: one containing a ~50 GB file generated with fallocate -l 50G 50g_file, and another containing 51 200 files of 1 MB each created via

seq 51200 | xargs -i dd if=/dev/zero of=1m_file{} bs=1M count=1

. The du command reported 51 GB due to filesystem block allocation.

Tool Tests

SCP

# 1 * 50G file test
[root@yang-02 big]# scp /opt/test/big/50g_file root@yang-01:/opt/test/re/
50g_file                          100%   50GB 135.5MB/s   06:17

# 51200 * 1M file test
[root@yang-02 many]# time scp /opt/test/many/1m_file* root@yang-01:/opt/test/re/
... (output omitted) ...
real    20m43.875s

SCP is widely available, uses SSH encryption, and shows moderate speed; CPU usage is low.

FTP

# 1 * 50G file test
[root@yang-01 re]# ftp yang-02
... (output omitted) ...
53687091200 bytes received in 150 secs (359 MB/s)

# 51200 * 1M file test
[root@yang-01 re]# time ftp yang-02
... (output omitted) ...
real    14m32.032s

FTP works over TCP, offers the fastest transfer for the large file but lacks encryption.

SFTP

# 1 * 50G file test
[root@yang-01 re]# sftp root@yang-02
... (output omitted) ...
100%   50GB 128.7MB/s   06:37
real    19m43.154s

SFTP adds SSH encryption; speed drops roughly 70 % compared with FTP, similar to SCP.

RSYNC

# 1 * 50G file test
[root@yang-02 big]# time rsync -av ./50g_file root@yang-01:/opt/test/re/50g_file
... (output omitted) ...
real    8m17.039s

# 51200 * 1M file test
[root@yang-02 many]# time rsync -av ./1m_file* root@yang-01:/opt/test/re/
... (output omitted) ...
real    15m21.548s

Rsync is slightly faster than SCP, supports compression (-z), and only transfers changed blocks, making it ideal for incremental backups.

TFTP

# 1 * 50G file test
[root@yang-01 re]# time tftp yang-02
... (output omitted) ...
real    10m30.114s (transfer failed, only 1.8 GB received)

TFTP uses UDP, requires additional service setup, and performed poorly for large files.

NC (netcat)

# 1 * 50G file test
[root@yang-02 big]# nc 192.168.88.71 10086 < /opt/test/big/50g_file
[root@yang-01 re]# time nc -l 10086 > 50G_file
real    2m30.663s

# 51200 * 1M file test
[root@yang-01 many]# tar cfz - *|nc 192.168.88.71 10086
... (output omitted) ...
real    11m38.400s

Netcat transfers data with minimal protocol overhead, achieving the fastest speeds among the tested tools.

Python SimpleHTTPServer

# 1 * 50G file test
[root@yang-02 big]# python -m SimpleHTTPServer 10086
... (output omitted) ...
wget http://192.168.88.72:10086/50g_file
... 330 MB/s ...

# 51200 * 1M file test
[root@yang-02 many]# python -m SimpleHTTPServer 10086
... (output omitted) ...
wget -i liu.list
... total 50G in 8m11s (104 MB/s) ...

The built‑in Python HTTP server provides a lightweight solution with good throughput and no extra installation.

Results Summary

Tool

50 GB Transfer Time

51 200 × 1 MB Transfer Time

Compression

Encryption

Speed

SCP

06:17 (377 s)

20m43.875s (1243.9 s)

-C supported

SSH

Fast

FTP

150 s

14m32.032s (872.0 s)

None

None

Fast

SFTP

06:37 (397 s)

19m43.154s (1183.2 s)

None

SSH

Fast

RSYNC

08:17.039s (657.0 s)

15m21.548s (921.5 s)

-z supported

SSH

Fast

TFTP

10m30.114s (test fail)

—

None

None

Slow

NC

02:30.663s (150.7 s)

11m38.400s (698.4 s)

None

None

Fast

Python HTTP

02:35 (155 s)

09:19 (559 s)

None

None

Fast

Conclusions

Large‑file transfer speed ranking: FTP > NC > Python HTTP > SCP > SFTP > RSYNC .

Small‑file transfer speed ranking: Python HTTP > NC > FTP > RSYNC > SFTP > SCP .

RSYNC excels in incremental synchronization and periodic archiving; FTP is simple but insecure and suited for internal networks.

When SSH is blocked, netcat (nc) provides a fast alternative using raw TCP/UDP.

Transferring many small files incurs higher CPU and I/O overhead compared with a single large file.

Parallelizing transfers (e.g., multiple processes) can better utilize bandwidth for massive file sets.

All tests were conducted on virtual machines; actual performance may vary with network bandwidth, disk I/O, and other environmental factors.