How We Test: SmallNetBuilder's Wireless Testbed - Revision 10

[thiggins]

The new SmallNetBuilder wireless testbed will handle both single point and distributed Wi-Fi router testing

Read on SmallNetBuilder

 

[eikido]

That test setup looks expensive, how could you afford it?

 

[thiggins]

Octoscope has generously provided the system on loan.

 

[eikido]

Sounds great That's really nice of them.

When will we see test results with Revision 10. It looks promising.

 

[thiggins]

I am working on first reviews now.

 

[eikido]

thanks mate! Your doing the world a favor

 

[Andy Tsai]

Hi Thiggins,
May I ask a several question for Revision 10:
1.According to your article , we know that v10 have replaced the Client M600 to PALs. If we try to used NETGEAR R7800 switch to client mode to instead the PALs setting , is that available? (NOTE: R7800 got the same solution with PALs)

2. According to How We Test: SmallNetBuilder's Wireless Testbed - Revision 10 ,paragraph "Computer", which said "we've reassigned the two Dell Optiplex SFF machines to wired router test duties ", what's that supposed to mean? That is, You set two Dell PCs for server (same subnet ex. Server1: 192.168.15.7 , Server2: 192.168.15.8), and connected to correspond ip address (like 192.168.15.7 to 192.168.15.9 ) ? Please let us clarify the setting if there has a block diagram, thanks.

thanks,

 

[thiggins]

1.According to your article , we know that v10 have replaced the Client M600 to PALs. If we try to used NETGEAR R7800 switch to client mode to instead the PALs setting , is that available? (NOTE: R7800 got the same solution with PALs)

I'm unclear on your question. Our standard test client is now octoScope's Pal device. You can use anything you want as a client device. But your results will not correlate with ours.

2. According to How We Test: SmallNetBuilder's Wireless Testbed - Revision 10 ,paragraph "Computer", which said "we've reassigned the two Dell Optiplex SFF machines to wired router test duties ", what's that supposed to mean? That is, You set two Dell PCs for server (same subnet ex. Server1: 192.168.15.7 , Server2: 192.168.15.8), and connected to correspond ip address (like 192.168.15.7 to 192.168.15.9 ) ? Please let us clarify the setting if there has a block diagram, thanks.

Wired router test process is described separately.
https://www.smallnetbuilder.com/lanwan/lanwan-howto/33107-how-we-test-hardware-routers-revision-10

 

[Andy Tsai]

I'm unclear on your question. Our standard test client is now octoScope's Pal device. You can use anything you want as a client device. But your results will not correlate with ours.


Wired router test process is described separately.
https://www.smallnetbuilder.com/lanwan/lanwan-howto/33107-how-we-test-hardware-routers-revision-10

thanks, mate!

 

[Easy Rhino]

Hey Tim I just had a thought about the peak 5ghz throughput tests.

These tests are done testing throughput to a PC with a wired 1 gigabit ethernet connection plugged into a LAN port, right?

For the fastest 4x4 routers, do you think that the gig ethernet might be limiting the peak speeds? Maybe it would be faster if both the client and the server were both on the 5ghz wifi (especially for those routers that have dual 5ghz bands).

Once all routers have 10gb ethernet ports it wouldn't be limiting any more

 

[coxhaus]

Can we do away from the ping test through a saturated router? That is just a QOS setting to give ping higher priority. We don't want to support DDOS attacks. In the old days telnet needed higher priority so if something went down we could still get into the device to work on it. Now days not so important as we move to web based. Old command line as in Cisco's IOS it is still important. But that is not tested here.

 

[thiggins]

Can we do away from the ping test through a saturated router? That is just a QOS setting to give ping higher priority. We don't want to support DDOS attacks. In the old days telnet needed higher priority so if something went down we could still get into the device to work on it. Now days not so important as we move to web based. Old command line as in Cisco's IOS it is still important. But that is not tested here.

Huh?

 

[coxhaus]

Your buffer bloat test. Where you ping through a saturated router.

 

[thiggins]

Your buffer bloat test. Where you ping through a saturated router.

So you don't want me to test for bufferbloat anymore? Or you want a different method used?

 

[coxhaus]

It seems like it should not be a QOS setting.

 

[sfx2000]

So you don't want me to test for bufferbloat anymore? Or you want a different method used?

@coxhaus - he makes a point - and the bufferbloat testing is always a bit odd with traffic patterns - with QoS - some traffic is going to be preferred over others, and this can skew results.

In any event - BB is most sensitive to asymmetric connections, esp. when the uplink BW is less than 1Mb/Sec - many customer connections these days are faster.

Just my thoughts here - and much of the bufferbloat is upstream, not in the router itself unless very carefully modeled and the problem understood.