Equipment Support > Hardware By Brand > ZyXEL > USG200 speed tests #2

USG200 speed tests #2

USG 200 2.12 netperf tests

tweety:~ # netperf -V
Netperf version 2.4.5
tweety:~ # netperf -f m -H 192.168.50.2
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.50.2 (192.168.50.2) port 0 AF_INET
 

Well done.

Would you please test the BMW Throughput under AppPatrol?

I couldn't digest the results so I ran another test this time using ftp protocol. I was downloading exactly the same 330MB zipped iso as in my first 2.11 set of tests »USG200 speed tests

FTP server was vsftpd and I was downloading with wget redirecting output to /dev/null to get rid of possible disk write delays.

Here is an example from just the firewall test

brano@tweety:~> wget -O /dev/null ftp://192.168.50.2/iso.zip
--2009-07-04 00:04:30--  ftp://192.168.50.2/iso.zip
           => `/dev/null'
Connecting to 192.168.50.2:21... connected.
Logging in as anonymous ... Logged in!
==> SYST ... done.    ==> PWD ... done.
==> TYPE I ... done.  ==> CWD not needed.
==> SIZE iso.zip ... 344458362
==> PASV ... done.    ==> RETR iso.zip ... done.
Length: 344458362 (329M)
 
100%[================================================================================== ==========>] 344,458,362 16.3M/s   in 20s
 
2009-07-04 00:04:51 (16.0 MB/s) - `/dev/null' saved [344458362]
 

USG200 2.12 ftp test

For IDP/AV performance, you can try to test it with multiple connections. The aggregate multiple session performance will be much better than single session performance.

BR,

style4me32

It probably would, but that's not the point of my test.
DUT should perform to advertised specs for single session.
--
openSUSE 11.1, KDE 4.2

See my #1 speed test with 2.11 firmware »USG200 speed tests
The performance then was mostly up to specs (excluding AppPatrol).
Something seems to be wrong in 2.12.
--
openSUSE 11.1, KDE 4.2

reply to Brano
Under the ftp test, does the USG200 has high CPU usage issue? What is the USG200 CPU usage when download the 330MB iso file?

I have a 30Mbps Internet connection (PPPoE) and using the ZyWALL USG200, when downloading using FTP the USG200 CPU usage is quite high 60%~80%.

When downloading a 1300MB ISO file using FlashGet and set to 8 simultaneous connections in microsoft.com, the USG200 CPU usage is >85%, does the USG200 has a PPPoE or CPU usage high issue?

Testing ISO File:
»www.microsoft.com/downloads/deta···3d06d6cb

reply to Brano
I've re-done the tests with the 2.12(AQU.1) firmware release.
...the results are not a bit better

reply to Brano
Hi Brano, based on your experience, if one has a 25/25 service and wanted to procure a USG router which one would you lean towards?
A. assume firewall plus AV plus IDP in use.

I did find some differences in available literature between US and UK sites both with 2009 data.

According to the spec sheet from zyxel USA for the USG200
Firewall / VPN / AV / IDP / AV + IDP + FIREWALL /
150 Mbps / 75 / 50 / 65 / 45 Mbps

and for the USG300
200 Mbps/ 100 / 70 / 75 / and 70 respectively.

I am assuming that those are all two way throughputs and for one-way we would have to divide all those numbers in half.

USG200 - 75 / 37.5 / 25 / 32.5 / 22.5
USG300 - 100 / 50 / 35 / 37.5 / 35

From the UK site a slightly different story ????»www.zyxel.co.uk/upload/doc/QSG_S···2009.pdf

There are no AV only or IDP only numbers for starters.
Firewall / VPN / AV + IDP + Firewall /
USG200 - 150 / 75 / 24
USG300 - 200 / 100 / 48

As you may surmise, it appears that the UK VPN and AV numbers are two way throughput and the combined AV + IDP + Firewall numbers are one way throughput. All very confusing????
--
Ain't nuthin but the blues! "Albert Collins".
Leave your troubles at the door! "Pepe Peregil" De Sevilla. Just Don't Wifi without WPA, "Yul Brenner"

LlamaWorks Equipment

Well, I have done only single session tests as per above and only on USG200. That said, AV speed seems to be the deal breaker for that test.

One would need additional tests, multi-session and single session tests each way to get the overall picture.

Solely based on ZyXel specs USG 200 should be able to handle 25/25 assuming they fix things in next firmware releases.
--
openSUSE 11.1, KDE 4.2

reply to Brano
I've been trying to diagnose a problem that has been bugging me for weeks. An application that we use on the network experiences sudden drop in performance, and the server logs show high spikes in network traffic (we're running gigabit). I think I've narrowed it down to the USG200. I'm running IDP+ADP+AV+CF, and I was running the 2.12 firmware. I've upgraded to the latest firmware 2.12(AQU.2), and I'm hoping it helps.

Brano, your speed tests have been very helpful. Have you had a chance to rerun the tests for the latest firmware 2.12(AQU.2)?

reply to Brano

Zyxel AV only.

Brano,

Could you re-run the test, but break the single 300Mb file up into multiple chunks of only 2-3Mb ?

I'd be interested to see if that improves AV performance (and it would be a more realistic real-life test I think). We've seen even a high-end Bluecoat AV appliance crap itself on "never-ending" streaming content (think stock-tickers and the like) so the extremely poor AV performance may very well be related to the very large file you're throwing at it.

Sorry, I'm extremely busy at work at the moment.
Next test I'm planning to do (IF my schedule allows) is 2.20. It's in beta 5 now so I'm hoping it won't take long until release.
--
When you do something, do it right!

reply to Eric_T

Doing multiple files simultaneously definitely adds to the quality of test.
The DUT (Device Under Test) may be performing differently on one session and multiple sessions. In theory it should deliver the advertised speeds regardless on number of sessions (often not true though).
I've chosen one file due to simplicity just to get "a picture". I don't have too much time to waste on this, but I was too curious to not do it

As for RAM-to-RAM tests and real file tests I've done both.
As you can see first set of tests is done using netperf (RAM-to-RAM) and the second is done using downloading real file via FTP.
To check whether the tests are limited by disk speeds I've done the first one without firewall to use for comparison. It's at the advertised 150Mb/s maximum. The real file download speed is equivalent to netperf speed. All subsequent tests are slower thus safe to conclude the slowdown is caused by DUT.

--
When you do something, do it right!