Are the WAN-Lan throughout results posted at SNB realistic?

[wayner]

I have an Asus RT-N66U router that according to the WAN-LAN throughput test on this site can handle 731.6Mbps. But my real world results show that when I use this router my WAN speed gets maxed out at about 230MBps (not using my router and connecting directly to my modem/gateway gets me 320Mbps).

My results are corroborated in this thread http://www.snbforums.com/threads/rt...expected-on-high-speed-fibre-connection.9425/ and apparently there are other similar threads it looks like the CPU is maxed out.

So where does the 731 result come from and is that a test in an idealized environment that is not useful in real world conditions!

My internet connection is Rogers Cables 250/20 service which actually gives download speeds of a bit over 300 Mbps. My gateway is a Hitron CGN3 and it is running in bridge mode.

 

[L&LD]

All of the tests presented in the charts are best used as a comparative number between two or more router models. As for WAN to LAN tests, they should be achievable when using a wired client.

Your issue on the thread you link is puzzling. But I think you need to update your year old firmware to get any help first as I have already suggested on that thread.

But your question here has been answered in the first two sentences.

 

[wayner]

But this problem isn't unique to me as you will have seen in the 2013 thread from the BT customer as others have reported the exact same issue. And one of the "solutions" is to roll back to an even older version of Merlin from 2013. Unless Merlin has done something to specifically address this issue then I am doubtful that upgrading will resolve this issue. If I am going to go through a reconfigure then I may be better off getting a new router with a faster CPU that is less likely to be prone to this issue.

In all instances my network test are run wired - I know that you won't get reliable results over wireless and I don't really use the wifi connection on my router very much.

But I don't think my question has been answered. Are the WAN-LAN throughput tests run with a typical ISP configuration that people would actually use? By that I mean using a cable or DSL modem or something like a Google Fiber modem on the WAN side? Or is the WAN side simulated?

If the WAN-LAN throughput result is 731.6 then shouldn't I be able to get that speed assuming my WAN speed is that fast? That certainly does not appear to be the case. It looks like you struggle to get one third of that rate.

And here is a thread that I started on DSLR that has other folks corroborating my results. http://www.dslreports.com/forum/r29909628-Anyone-using-Rogers-250-with-an-Asus-RT-N66U-

 

[L&LD]

The question was answered.

You simply seem to be stuck and not willing to test much newer, better and more secure firmware.

You don't really think RMerlin will support old firmware that has been depreciated from a year ago?

Even if you do nothing else, I would at least reset to factory defaults and manually and minimally setup the router to connect to your ISP (at this point, I would unplug the modem for at least 30 minutes too, before booting it up and connecting the router again) and secure your router and wifi ssids. This is with the same firmware you have now.

But, if you go to all that trouble... may as well get your router current too.


In that thread you link,

http://www.dslreports.com/forum/r29909684-

gw280 states he hit 325 down with a much more current version of the stock firmware than what you are running. And RMerlin is based on the stock Asus firmware, as you know.

Seems pretty clear to me. Update your firmware.

 

[Nullity]

I think the reported speeds are the "forwarding" speeds, not the routing speeds.

You need to have another device making the routing decisions and leave the RT-N66U to only perform forwarding. I believe this is hows you exploit the "hardware acceleration". The hardware acceleration does not affect routing speeds, afaik.

 

[L&LD]

You've mentioned this a few times now.

Isn't routing, forwarding?

 

[wayner]

According to shibby20 (I believe he is the author of the Tomato firmware) in this thread the max WAN-LAN throughput of the RT-N66U is 240 Mbps wihtout bcm_nat enabled and is 330 with bcm_nat enabled. Either is a long way from 731.

http://www.linksysinfo.org/index.ph...age-for-wan-lan-throughput.70151/#post-247632

 

[Nullity]

You've mentioned this a few times now.

Isn't routing, forwarding?

Yes... It was a poor attempt to explain how Broadcom's CTF module may work.

Regardless of my incorrect assumptions, a sizeable minority have mentioned achieving 700-900Mbps speeds with the RT-N66U. I personally get ~300Mbps with CTF/hardware-acceleration disabled.

Apologies for my confusing, poor quality posts regarding routing/forwarding and CTF.

 

[thiggins]

The router test process is described here: http://www.smallnetbuilder.com/lanwan/lanwan-howto/31103-how-we-test-hardware-routers-revision-3

The only things "idealized" about the test is that it uses a DHCP connection (not PPPoE / PPTP, etc) and the IxChariot scripts may be more efficient at sending traffic than normal network connections.

 

[taoyang]

thiggins

Could you add out-of-the-box firewall throughput speeds to your reviews ?

So the charts have some meaning of comparison for everyday use as most home users are going to be running it enabled ?

Thanks

 

[thiggins]

thiggins

Could you add out-of-the-box firewall throughput speeds to your reviews ?

So the charts have some meaning of comparison for everyday use as most home users are going to be running it enabled ?

I don't understand your request. All routers are tested with factory default, which usually means the NAT firewall is enabled.

 

[htismaqe]

You've mentioned this a few times now.

Isn't routing, forwarding?

Forwarding also occurs at the datalink layer. For example, all ethernet switches have a forwarding table full of MAC addresses that are being switched (or broadcast) rather than routed. Sorry, couldn't help myself.

 

[taoyang]

I don't understand your request. All routers are tested with factory default, which usually means the NAT firewall is enabled.

As far as smallnetbuilder, their "benchmarks" if I dare call them that, are worthless. Those guys have a long way to go before they start producing anything close to being a benchmark. Only a fool would fall for a benchmark of a security appliance that gives numbers based solely on disabling every piece of security measure on it. They might as well just test the lan switching speed. They provide in-depth overviews--and not much more.

http://community.ubnt.com/t5/EdgeMAX/The-new-EdgeRouter-X-and-EdgeRouter-X-SFP/td-p/1222426/page/13

My Bad , Sorry.

It was suggested in the above thread that you disable security to get your higher speeds throughputs for the charts.

 

[Nullity]

Forwarding also occurs at the datalink layer. For example, all ethernet switches have a forwarding table full of MAC addresses that are being switched (or broadcast) rather than routed. Sorry, couldn't help myself.

Exactly. My knowledge of computer networking was too sparse then to properly articulate.

I still have no idea how to achieve those speeds with my RT-N66U though... I guess LAN-LAN is different. *shrug*

 

[thiggins]

It was suggested in the above thread that you disable security to get your higher speeds throughputs for the charts.

We try to clearly describe all of our test methods. Router test method is
http://www.smallnetbuilder.com/lanwan/lanwan-howto/31103-how-we-test-hardware-routers-revision-3

Many routers today use cut-through-forwarding to achieve near Gigabit wire speed routing. This is usually enabled by default.

If the manufacturer provides a disable, we usually also test routing performance with CTF disabled. This always results in LOWER throughput. But the data posted in the charts is always with default settings.

 

[htismaqe]

Exactly. My knowledge of computer networking was too sparse then to properly articulate.

I still have no idea how to achieve those speeds with my RT-N66U though... I guess LAN-LAN is different. *shrug*

Unless you REALLY need it, there's no reason to turn it on. It bypasses a lot of arguably essential parts of the protocol (e.g. it forwards ethernet frames prior to the CRC being received, so it could be forwarding errors instead of valid frames).

 

[System Error Message]

Im sure testing was done on the router's NAT throughput and many consumer routers use hardware acceleration.

When it comes to switching it doesnt matter what method of switching was used. In terms of just forwarding traffic both store and forward and cut through forwarding(slightly lower latency) have the same throughput however throughput drops in L2 forwarding in store and forward if other features are applied but this only applies to LAN switching and isnt relevant to routing or NAT speeds. Since consumer routers do not have the option to configure their switch as managed switches can therefore the type of switching used in these routers are a moot point.

From my experience using a semi managed store and forward switch the error checking capability isnt necessary because there isnt any packet drops in a normal ethernet network. All applications now work on L3 at least so unless you play games that make use of IPX or L2 networking or have devices of different ethernet speeds or having traffic convert between networking mediums than there isnt much use of store and forward except for the features it provides.

 

[taoyang]

We try to clearly describe all of our test methods. Router test method is
http://www.smallnetbuilder.com/lanwan/lanwan-howto/31103-how-we-test-hardware-routers-revision-3

Many routers today use cut-through-forwarding to achieve near Gigabit wire speed routing. This is usually enabled by default.

If the manufacturer provides a disable, we usually also test routing performance with CTF disabled. This always results in LOWER throughput. But the data posted in the charts is always with default settings.

Thanks

Shame the UBNT forums are not so helpful.

 

[Nullity]

Unless you REALLY need it, there's no reason to turn it on. It bypasses a lot of arguably essential parts of the protocol (e.g. it forwards ethernet frames prior to the CRC being received, so it could be forwarding errors instead of valid frames).

Oh right. I forgot that things like tcpdump and QoS break with CTF enabled.

 

[htismaqe]

Right. With most manufacturers like Asus and Netgear, enabling anything that requires deeper packet inspection than source/destination pair disables CTF.