PDA

View Full Version : ioFTPD SSL Benchmark Tests


Mouton
11-04-2003, 02:44 PM
Machine:
AMD Athlon 1100Mhz
1GB PC133 RAM
Windows 2003 Advanced
On-board NIC (SiS 900)

Note: Pretty crappy NIC, I know, but the goal is just to see the speed difference between SSL and no SSL transfers, not measure the transfer speed.

Note 2: Added 40 bits encryption, which shows in FlashFXP as RC4 128 bits.

Results:
http://mouton.abuserz.com/tests_ssl/results.htm

During SSL Transfers, CPU was maxed out by FlashFXP and ioFTPD:
http://mouton.abuserz.com/tests_ssl/performance.gif

Raw data:
http://mouton.abuserz.com/tests_ssl/results.txt
http://mouton.abuserz.com/tests_ssl/results.xls

Conclusion: SSL transfers using 128 bits encryption go about 53% of the speed of no SSL transfers.

kraxdaz
11-04-2003, 03:11 PM
thanks for this test mouton !!:cool:

Linkster
11-04-2003, 04:34 PM
so the client and server were on the same machine? a more accurate result would be to test with a true client connected via crossover or a good switch...the fact that the server has to do both the encoding and decoding skews the results and does not represent a real world scenario.

bounty
11-04-2003, 05:01 PM
Conclusion: if you want performances .... forget SSL ;)

big thx for these tests

bounty

Mouton
11-04-2003, 06:58 PM
Linkster: u're welcome to do those tests and report the results you get.

Linkster
11-04-2003, 10:54 PM
I knew you were going to say that :p

I'll see what i can do when i can get around to it...they probably won't be as detailed and nice as yours though :)

odd
11-07-2003, 01:15 PM
When Ive saw Mouton's results ive decided to benchmark myself but this
time to a dedicated ioFTPD server running with 1,2ghz Duron, 768mb DDR,
7200rpm disks, KT400 Mainboard, D-link 10/100 NIC. I tested both upload
and download but there wasnt any difference in the speeds.I think there
is a big difference between my result and Mouton.


Require_Encrypted_Auth = *
Require_Encrypted_Data = *
Certificate_Name = xxxx
Explicit_Encryption = True
Encryption_Protocol = SSL3
Min_Cipher_Strength = 128
Max_Cipher_Strength = 384


Here ive tested 15mb file.
__Without_encryption__
150 Opening BINARY mode data connection for some.sort.of.file.rar.
226 [PWD: /archive/some.sort.of.file/ ]-[Section: Default]-[Credits: 2.9M]-[Ratio: Unlimited]-[Speed: 8200.1kb/s]
Transferred: some.sort.of.file.rar. 15_000_000 bytes in 1,83 (8_013,37 KBps)

Transfer queue completed
Transferred 55 files totaling 747_034_460 bytes in 01:16 (9_499,40 KBps)


__With_128bit_encryption__
Negotiating SSL/TLS session...
150 Opening BINARY mode data connection for some.sort.of.file.rar.
SSL/TLS negotiation successful...
SSL/TLS connection using cipher RC4-MD5 (128 bits)
226 [PWD: /archive/some.sort.of.file/ ]-[Section: Default]-[Credits: 2.9M]-[Ratio: Unlimited]-[Speed: 6277.5kb/s]
Transferred: some.sort.of.file.rar 15_000_000 bytes in 2,39 (6_126,49 KBps)

Transfer queue completed
Transferred 55 files totaling 747_034_460 bytes in 02:04 (6_017,40 KBps)

Here ive tested a 50mb file
___Without_encryption__
150 Opening BINARY mode data connection for 50mb.file3.rar.
226 [PWD: /Upload/ ]-[Section: Default]-[Credits: 2.9M]-[Ratio: Unlimited]-[Speed: 10491.2kb/s]
Transferred: 50mb.file3.rar 52_500_000 bytes in 5,00 (10_253,91 KBps)

Transfer queue completed
Transferred 3 files totaling 157_500_000 bytes in 15,28 (10_200,86 KBps)


__With_128bit_encryption__
Negotiating SSL/TLS session...
150 Opening BINARY mode data connection for 50mb.file3.rar.
SSL/TLS negotiation successful...
SSL/TLS connection using cipher RC4-MD5 (128 bits)
226 [PWD: /Upload/ ]-[Section: Default]-[Credits: 2.9M]-[Ratio: Unlimited]-[Speed: 6259.5kb/s]
Transferred: 50mb.file3.rar 52_500_000 bytes in 8,25 (6_214,49 KBps)

Transfer queue completed
Transferred 3 files totaling 157_500_000 bytes in 25,05 (6_214,49 KBps)
I might have done something wrong but I dont think so. I wasnt sure what Mouton got and his url is down. Ive tested to send from myself to myself with encryption(128bit) and got around 3mb/s.

darkone
11-07-2003, 01:31 PM
I think he somehow bounced it through network.. so card was actually sending/receiving data which causes cpu load increase on cheaper cards