dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
10096
share rss forum feed

cooldude9919

join:2000-05-29
kudos:5

1921 vs 891 Throughput Testing

Figures i got this completed the day before the site went down, i know there where some of you waiting on this.

Not as impressed with the 1921 as i hoped i would be when compared to the 891, but it is still better. If the 891 had 2 gig ports instead of one it would be capable of more.

 
1921 NAT ONLY
CPU utilization for five seconds: 92%/91%;
15.0-1 m8
 
C:\>iperf -c 10.1.18.95 -P 3 -t 60
------------------------------------------------------------
Client connecting to 10.1.18.95, TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[1880] local 10.3.152.2 port 1829 connected with 10.1.18.95 port 5001
[1912] local 10.3.152.2 port 1827 connected with 10.1.18.95 port 5001
[1896] local 10.3.152.2 port 1828 connected with 10.1.18.95 port 5001
[ ID] Interval       Transfer     Bandwidth
[1896]  0.0-60.0 sec   866 MBytes   121 Mbits/sec
[1912]  0.0-60.0 sec   832 MBytes   116 Mbits/sec
[1880]  0.0-60.0 sec   824 MBytes   115 Mbits/sec
[SUM]  0.0-60.0 sec  2.46 GBytes   352 Mbits/sec
 
CISCO 891 NAT ONLY
flash:c890-universalk9-mz.150-1.M6.bin
CPU utilization for five seconds: 21%/21%
C:\>iperf -c 10.1.18.95 -P 3 -t 60
------------------------------------------------------------
Client connecting to 10.1.18.95, TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[1912] local 10.3.152.2 port 1606 connected with 10.1.18.95 port 5001
[1880] local 10.3.152.2 port 1608 connected with 10.1.18.95 port 5001
[1896] local 10.3.152.2 port 1607 connected with 10.1.18.95 port 5001
[ ID] Interval       Transfer     Bandwidth
[1880]  0.0-60.0 sec   219 MBytes  30.6 Mbits/sec
[1912]  0.0-60.0 sec   225 MBytes  31.5 Mbits/sec
[1896]  0.0-60.0 sec   225 MBytes  31.5 Mbits/sec
[SUM]  0.0-60.0 sec   670 MBytes  93.6 Mbits/sec
 
1921 ZBFW & NAT
CPU utilization for five seconds: 92%/92%
C:\>iperf -c 10.1.18.95 -P 3 -t 20
------------------------------------------------------------
Client connecting to 10.1.18.95, TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[1880] local 10.3.152.2 port 1865 connected with 10.1.18.95 port 5001
[1896] local 10.3.152.2 port 1864 connected with 10.1.18.95 port 5001
[1912] local 10.3.152.2 port 1862 connected with 10.1.18.95 port 5001
[ ID] Interval       Transfer     Bandwidth
[1896]  0.0-20.0 sec   133 MBytes  56.0 Mbits/sec
[1880]  0.0-20.0 sec   131 MBytes  55.0 Mbits/sec
[1912]  0.0-20.0 sec   133 MBytes  55.6 Mbits/sec
[SUM]  0.0-20.0 sec   397 MBytes   167 Mbits/sec
 
891 ZBFW & NAT
CPU utilization for five seconds: 99%/83%
 
C:\>iperf -c 10.1.18.95 -P 3 -t 20
------------------------------------------------------------
Client connecting to 10.1.18.95, TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[1912] local 10.3.152.2 port 1998 connected with 10.1.18.95 port 5001
[1880] local 10.3.152.2 port 2000 connected with 10.1.18.95 port 5001
[1896] local 10.3.152.2 port 1999 connected with 10.1.18.95 port 5001
[ ID] Interval       Transfer     Bandwidth
[1912]  0.0-20.0 sec   107 MBytes  44.8 Mbits/sec
[1896]  0.0-20.0 sec   101 MBytes  42.3 Mbits/sec
[1880]  0.0-20.0 sec  13.5 MBytes  5.66 Mbits/sec
[SUM]  0.0-20.0 sec   221 MBytes  92.8 Mbits/sec
 
1921 ZBFW ONLY NO NAT
CPU utilization for five seconds: 93%/92%
C:\>iperf -c 10.1.18.95 -P 3 -t 20
------------------------------------------------------------
Client connecting to 10.1.18.95, TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[1912] local 10.3.152.2 port 1883 connected with 10.1.18.95 port 5001
[1896] local 10.3.152.2 port 1884 connected with 10.1.18.95 port 5001
[1880] local 10.3.152.2 port 1885 connected with 10.1.18.95 port 5001
[ ID] Interval       Transfer     Bandwidth
[1880]  0.0-20.0 sec   239 MBytes   100 Mbits/sec
[1896]  0.0-20.0 sec   233 MBytes  97.8 Mbits/sec
[1912]  0.0-20.0 sec   240 MBytes   100 Mbits/sec
[SUM]  0.0-20.0 sec   712 MBytes   299 Mbits/sec
 
891 ZBFW ONLY NO NAT
yourname#sh proc cpu
CPU utilization for five seconds: 26%/25%
 
C:\>iperf -c 10.1.18.95 -P 3 -t 30
------------------------------------------------------------
Client connecting to 10.1.18.95, TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[1880] local 10.3.152.2 port 2003 connected with 10.1.18.95 port 5001
[1896] local 10.3.152.2 port 2002 connected with 10.1.18.95 port 5001
[1912] local 10.3.152.2 port 2001 connected with 10.1.18.95 port 5001
[ ID] Interval       Transfer     Bandwidth
[1896]  0.0-30.0 sec   112 MBytes  31.2 Mbits/sec
[1880]  0.0-30.0 sec   112 MBytes  31.4 Mbits/sec
[1912]  0.0-30.0 sec   110 MBytes  30.9 Mbits/sec
[SUM]  0.0-30.0 sec   334 MBytes  93.5 Mbits/sec
 
 

badsykes1

join:2004-12-08

I don't know what ZBFW is ,from the name feels like a firewall , but it seems to me that the 891 is optimized for this

891 ZBFW ONLY NO NAT
yourname#sh proc cpu
CPU utilization for five seconds: 26%/25%

1921 ZBFW ONLY NO NAT
CPU utilization for five seconds: 93%/92%

Also please explain to me this values:
1921 NAT ONLY
CPU utilization for five seconds: 92%/91%;
15.0-1 m8

[1896] 0.0-60.0 sec 866 MBytes 121 Mbits/sec
[1912] 0.0-60.0 sec 832 MBytes 116 Mbits/sec
[1880] 0.0-60.0 sec 824 MBytes 115 Mbits/sec
[SUM] 0.0-60.0 sec 2.46 GBytes 352 Mbits/sec

CISCO 891 NAT ONLY
flash:c890-universalk9-mz.150-1.M6.bin
CPU utilization for five seconds: 21%/21%
[1880] 0.0-60.0 sec 219 MBytes 30.6 Mbits/sec
[1912] 0.0-60.0 sec 225 MBytes 31.5 Mbits/sec
[1896] 0.0-60.0 sec 225 MBytes 31.5 Mbits/sec
[SUM] 0.0-60.0 sec 670 MBytes 93.6 Mbits/sec

They feel weird...
Why 219 MBytes (the size of the file delivered only 30mbit and the 670MB file delivered 93.6Mbits (the full 100mbit port capacity)?

Also same problem for 1921 ...2.46GB file 352mbit ? --- gbit ? and with the an 866MB file got only 121Mbit ....

sorry if those are dumb question... i am not too much in Enterprise sector of equipment..


cooldude9919

join:2000-05-29
kudos:5

ZBFW is zone based firewall, so yes its a firewall. 891 isnt necessarily optimized moreso than the 1921, but i am suprised how well it does. If the 891 had 2x1gb it could pretty much keep up with the 1921 with zbfw throughput

I used IPERF for the throughput testing. I did 3 streams for 60 seconds, so it adds up how much data each stream had in the 60 seconds for the total data transfered and total throughput which is the last line of each section starting with [SUM].

1921 3 streams, each stream did what was listed, comes up to 2.46Gbyte which comes out to 352mbit over 60 seconds.

891 3 streams, each stream did what aws listed, comes up to 670Mbytes which comes out to 93mbit over 60 seconds, obviously the 100mbit port is the limit here not cpu.

hope this helps clear things up


HELLFIRE
Premium
join:2009-11-25
kudos:15
reply to cooldude9919

What's telling is the higher CPU util on the 1921 for NAT only versus the 891 for NAT only.
And I'm somewhat disheartened the 1921 can't reach .5 gigabit speeds. Guess that's Cisco's
incentive for you to buy the 29xx and 39xx series gear.

So what's the max thruput in a straight up network to network transfer cooldude9919? ie.
from 192.16.x.x/24 to 172.16.x.x/12 for example.

Regards



Da Geek Kid

join:2003-10-11
::1
kudos:1
reply to cooldude9919

1st of all, I recommend not to be on a 15 code unless it's mandatory like the 1900s...

2nd, nothing below 15.1 is stable enough to be considered Beta or I like to call Alpha code...


sk1939
Premium
join:2010-10-23
Mclean, VA
kudos:10
Reviews:
·T-Mobile US
·Verizon FiOS

said by Da Geek Kid:

1st of all, I recommend not to be on a 15 code unless it's mandatory like the 1900s...

2nd, nothing below 15.1 is stable enough to be considered Beta or I like to call Alpha code...

Interesting that you mention that. I don't care for 15 series code, but all of the X9XX generation devices don't have a 12.X iteration of code available to them, if I do recall.

HELLFIRE
Premium
join:2009-11-25
kudos:15
reply to cooldude9919

Question cooldude9919, switching back to a CBAC rather than ZBFW config, does it have much of a performance delta?

Regards


cooldude9919

join:2000-05-29
kudos:5
reply to cooldude9919

Here are some real world production tests from me. Both had IPS and zbfw on the interface. 1921 looks to be around ~60-70% faster in these tests.

891 router
 
NAT  60% CPU
30 second input rate 37836000 bits/sec, 3142 packets/sec
  30 second output rate 723000 bits/sec, 1582 packets/sec
 
No NAT
F0030001#sh proc cpu
CPU utilization for five seconds: 55%/53%; one minute: 39%; five minutes: 24%
 30 second input rate 35795000 bits/sec, 3163 packets/sec
  30 second output rate 500000 bits/sec, 939 packets/sec
 
1921 router
 
NAT
F0030001#sh proc cpu
CPU utilization for five seconds: 37%/36%; one minute: 24%; five minutes: 13%
 
 30 second input rate 36010000 bits/sec, 3252 packets/sec
  30 second output rate 1011000 bits/sec, 1815 packets/sec
     499657 packets input, 615665550 bytes, 0 no buffer
 
No NAT
 
F0030001#sh proc cpu
CPU utilization for five seconds: 32%/31%; one minute: 30%; five minutes: 32%
 PID Runtime(ms)     Invoked      uSecs   5Sec   1Min   5Min TTY Process
 
30 second input rate 37019000 bits/sec, 3339 packets/sec
  30 second output rate 893000 bits/sec, 1094 packets/sec
 
 

cramer
Premium
join:2007-04-10
Raleigh, NC
kudos:8
reply to cooldude9919

One-arm both of them... vlan tag the gige port and test vlan-to-vlan. That's as close to apples-to-apples as you'll be able to get. The 891 only *looks* better because it's limited to 100mbps by the available ports.