Bug 856336

Summary: 3des ipsec performance regression
Product: [Fedora] Fedora Reporter: Sergei LITVINENKO <sergei.litvinenko>
Component: kernelAssignee: Kernel Maintainer List <kernel-maint>
Status: CLOSED NOTABUG QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: 16CC: gansalmon, itamar, jonathan, kernel-maint, madhu.chinakonda
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2012-09-11 17:25:50 EDT Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---
Attachments:
Description Flags
sysprof saved profiles for 3des-sha1 and aes256 none

Description Sergei LITVINENKO 2012-09-11 15:50:54 EDT
Created attachment 611903 [details]
sysprof saved profiles for 3des-sha1 and aes256

Description of problem:
3des-sha1-96 encrypting privide 5 times less productivity as ase256.

Version-Release number of selected component (if applicable):
3.4.9-2.fc16.i686.PAE

How reproducible:
100%

Steps to Reproduce:
1. install and setup openswan
2. prepare 3des-sha1-96 configuration for 2 hosts
3. prepare ase256-sha configuration for the same 2 hosts 
4. connect 2 hosts by 1G ethernet
5. test network productivity by iperf (no_vpn/ 3des / aes256)
  
Actual results:
=============================
==> no VPN ( clean ethernet)
=============================
[root@ua-dudn00000 ~]# iperf -s -p 65000
------------------------------------------------------------
Server listening on TCP port 65000
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
[  4] local 10.x.x.104 port 65000 connected with 10.x.x.100 port 39617
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-100.0 sec  7.15 GBytes   614 Mbits/sec
[  5] local 10.x.x.104 port 65000 connected with 10.x.x.100 port 39618
[  5]  0.0-100.1 sec  7.14 GBytes   613 Mbits/sec

===============
== > 3des-sha1
===============
[root@homedesk ipsec.d]# iperf -c 10.x.x.104 -p 65000 -t 100
------------------------------------------------------------
Client connecting to 10.x.x.104, TCP port 65000
TCP window size: 21.0 KByte (default)
------------------------------------------------------------
[  3] local 10.x.x.100 port 39620 connected with 10.x.x.104 port 65000
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-100.1 sec   636 MBytes  53.4 Mbits/sec

================
==> aes256-sha1
================
[root@homedesk ipsec.d]# iperf -c 10.x.x.104 -p 65000 -t 100
------------------------------------------------------------
Client connecting to 10.x.x.104, TCP port 65000
TCP window size: 21.0 KByte (default)
------------------------------------------------------------
[  3] local 10.x.x.100 port 39621 connected with 10.x.x.104 port 65000
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-100.0 sec  2.90 GBytes   249 Mbits/sec

Expected results:

3des have to be not so much slowest as aes256.

Additional info:

PC for test : 
--------------
1. CPU: E6850  @3.00GHz
2. RAM: 6Go
Comment 1 Josh Boyer 2012-09-11 17:25:50 EDT
This is not going to be fixed in F16, at least not directly.  It's not even clear from the report that it is an actual bug or a regression.

Differing crypto algorithm implementations are going to have various performance impacts.  If you feel 3des is too slow, perhaps you could work with the upstream crypto maintainers to improve it.