WAN to LAN throughput testing differences + feedback

[wye]

Tim, I believe your WAN to LAN testing methodology is wrong.

In your N66U review you got:
WAN to LAN: 732 Mbps
And a simple SpeedTest that I did right now on my N66U contradicts you severely:

I got even 931 Mbps, which demolishes your router chart. Oh, and btw this is on PPPoE - with all that encoding overhead jazz.
For practical reasons, here is another example: I got 902 Mbps WAN to LAN in uTorrent with 25 peers. There is definitely something wrong with your benchmarking methodology/tools.

Also, while I'm here, here is another piece of advice: step down on your high-and-mighty attitude, you do have a lot to learn and its preventing you from doing a good job.

E.g. 1: in your N66U review, you say "Nothing's broken, folks. It's more throughput than anyone will ever use." Actually, I'm considering to use it for dual gigabit WAN links. And I plan to use 10 GbE for LAN at home. Which will be painfully slow for my 5-10 GIGABYTES/sec PCIe-SSDs. So my real needs are 100 GbE. Pretty far from 800 Mbits, ain't it?

E.g. 2: your AC87U review is a disgrace, you simply can't hold yourself from spilling never-ending bile towards Asus. We get it, you don't like Asus. Maybe you are right. But try to only say it once, not every second paragraph. Its extremely hard to find useful information from what pile of Asus-hating manure. Be more professional/objective. Stick.to.the.numbers. And make sure those numbers are good. Speak with other people, learn how they are doing their own network testing. Learn. Be a better person.

 

[System Error Message]

i think he hates ASUS because of the price and how ASUS locked down their firmwares despite region, basically their attitude towards their customers who bought it.

However if you look at wan to lan simultaneous there isnt yet a consumer router that exceeds 2 Gb/s. My router can do NAT at wirespeed and thats many gigabits. Not sure if he is willing to shell out the cash to get one lol. The brand i use is coming out with a new router soon that would have a total 160Gb/s of routing capacity. It would fulfill his 100Gb/s requirement but i think he would probably say he cant afford it despite it being much much cheaper than any consumer router for price/NAT performance. The only killer would be the price of those SFP+ modules and the SFP+ cards for his servers/desktops.


PPPOE is a layer 2 protocol so the overhead isnt as much as you would think comparing to the other protocols on layer2 but PPPOE does require some processing although routers can do more PPPOE than they can do VPN.

 

[sinshiva]

Which will be painfully slow for my 5-10 GIGABYTES/sec PCIe-SSDs. So my real needs are 100 GbE. Pretty far from 800 Mbits, ain't it?

which SSDs were those? something like; http://rog.asus.com/361862014/rampa...sd-raid-0-performance-on-rampage-v-extreme/2/ ?

can a non-commercial entity even attain more than a couple gigabit WAN links anywhere in the world?

 

[jegesq]

Tim, I believe your WAN to LAN testing methodology is wrong.

In your N66U review you got:
WAN to LAN: 732 Mbps
And a simple SpeedTest that I did right now on my N66U contradicts you severely:

I got even 931 Mbps, which demolishes your router chart. Oh, and btw this is on PPPoE - with all that encoding overhead jazz.
For practical reasons, here is another example: I got 902 Mbps WAN to LAN in uTorrent with 25 peers. There is definitely something wrong with your benchmarking methodology/tools.

Also, while I'm here, here is another piece of advice: step down on your high-and-mighty attitude, you do have a lot to learn and its preventing you from doing a good job.

E.g. 1: in your N66U review, you say "Nothing's broken, folks. It's more throughput than anyone will ever use." Actually, I'm considering to use it for dual gigabit WAN links. And I plan to use 10 GbE for LAN at home. Which will be painfully slow for my 5-10 GIGABYTES/sec PCIe-SSDs. So my real needs are 100 GbE. Pretty far from 800 Mbits, ain't it?

E.g. 2: your AC87U review is a disgrace, you simply can't hold yourself from spilling never-ending bile towards Asus. We get it, you don't like Asus. Maybe you are right. But try to only say it once, not every second paragraph. Its extremely hard to find useful information from what pile of Asus-hating manure. Be more professional/objective. Stick.to.the.numbers. And make sure those numbers are good. Speak with other people, learn how they are doing their own network testing. Learn. Be a better person.

I would trust Tim Higgin's testing methods for measuring WAN to LAN throughput (described here for the N66U) far more than I would a test which measures only internet download speed using an online source like the Ookla engine, which can and often is wildly inaccurate and a very crude measure at best. It also isn't measuring WAN to LAN throughput. You could learn a thing or two also from Tim by reading some of the articles at SNB on the basics, including the differences between speed and throughput (See, e.g., http://www.smallnetbuilder.com/wireless/wireless-basics/32175-how-fast-can-your-wi-fi-go).

Also, I don't know where you get the impression that Tim hates Asus. Now Mr. System Error Message, I certainly get that YOU hate Asus, based on precisely the criteria you ascribe to Tim (price and the lockdown of the router's country codes and channels). I think Tim has expressed the view repeatedly that he thinks the 87U is, at the moment, overpriced, and I think he's said the same things about the Linksys and Netgear 3200AC products as well, i.e., that for what you are getting, they are overpriced. I have read all of his Asus reviews carefully, and they seem to me to be fair, balanced, and based solely on the numbers from testing that is done with defined protocols and methods that are clearly described and repeatable, and his opinions are reserved for things like aesthetic placement of LED's, ports, vents, antennae, etc. System Error Message doesn't cite any review in particular, but Wye takes issue with his N66U review, which of course was written almost three years ago, when the N66U was first released and with the 87U review as well. I don't see the same negative slant that Wye notes in either review; in fact, Tim's review of the N66U seems very balanced where he concluded:

"clearly outperformed both the NETGEAR WNDR4500 and Cisco Linksys E4200V2 in most of our two and three-stream tests. And it's the only router in recent memory able to reach to our worst-case/lowest-signal test location on the 5 GHz band, albeit with barely-usable throughput. Still, this is an accomplishment in itself.

As for any reservations about documentation or firmware of the N66U when it was first issued, well, that router wasn't a finished product when Asus first released it and we all know that looking back over three years of history. The same can be said for the AC66U, the AC68U, and now the 87U, which is following the same trend (having the bugs worked out of the firmware over time). Tim's 87U review told it like it was at the time it was written: The hardware was intriguing, but the documentation was incomplete and spotty for many of the new feature sets (e.g., Adaptive QOS, etc), and there were as we know, a ton of bugs that needed to be worked out, including some with the 5ghz band that are still being worked out for many users. Oh, and when someone finally comes out with a 4x4 client (still waiting....tick-tock), maybe then we can all really know just how fast "fast" can be and how good the 87U can be.

But as for the "Tim-hate," I think it's enormously out of whack with what can objectively be found at this site, and if anything, I'd say that SNB gives almost more love and devotion to Asus than almost any other brand, and certainly more so than any other non-Asus-branded enthusiast site on the internet that I've seen. I mean, how else do you explain why there is not just a single forum devoted to Asus products, but three separate sub-forums exclusively about various Asus wireless products and firmware? Oh, yeah, could be because they are the only company that doesn't orphan users with planned obsolescence and instead continue to improve and support firmware many years after a product comes to market and long after newer and snazzier technology is released, and because there is a huge and avid user-base, including folks like Merlin, John and others who work on and develop independent firmware that make these products some of the best routers for SOHO wireless router enthusiasts.

Hey, we should all resolve to be better human beings and nicer people all the time. That certainly would include you as well, Wye.

 

[jegesq]

Yes,

Originally Posted by wye
Which will be painfully slow for my 5-10 GIGABYTES/sec PCIe-SSDs. So my real needs are 100 GbE. Pretty far from 800 Mbits, ain't it?

As of October 1, 2014, the fastest enterprise class SSD's currently on the market were:

PCIe SSDs Comparison Matrix

PCIe SSD Read GB/s Write GB/s

Dell PowerEdge Express NVMe 3 1.4
HGST s1122 1.4 1.1
Micron P420M 3.3 0.630
OCZ Z-Drive 4500 2.9 2.2
SanDisk / Fusion-io PX600 2.7 1.5

Hmm. I guess that's pretty far from your magical "5-10 Gigabytes/sec" SSDs.

Where do you get one of those SSD's you're packin?

Oh, but don't let facts get in your way while you snidely insult the site's host/reviewer.

 

[System Error Message]

I would trust Tim Higgin's testing methods for measuring WAN to LAN throughput (described here for the N66U) far more than I would a test which measures download using an online source from Ookla, which can and often is wildly inaccurate and a very crude measure at best.

Also, I don't know where you guys get the impression that Tim hates Asus. Now Mr. System Error Message, I certainly get that YOU hate Asus, based on precisely the criteria you ascribe to Tim (price and the lockdown of the router's country codes and channels). I think Tim has expressed the view repeatedly that he thinks the 87U is, at the moment, overpriced, and I think he's said the same things about the Linksys and Netgear 3200AC products as well, i.e., that for what you are getting, they are overpriced. I have read all of his Asus reviews carefully, and they seem to me to be fair, balanced, and based solely on the numbers from testing that is done with defined protocols and methods that are clearly described and repeatable, and his opinions are reserved for things like aesthetic placement of LED's, ports, vents, antennae, etc. System Error Message doesn't cite any review in particular, but Wye takes issue with his N66U review, which of course was written almost three years ago, when the N66U was first released and with the 87U review as well. I don't see the same negative slant that Wye notes in either review; in fact, Tim's review of the N66U seems very balanced where he concluded:



As for any reservations about documentation or firmware of the N66U when it was first issued, well, that router wasn't a finished product when Asus first released it and we all know that looking back over three years of history. The same can be said for the AC66U, the AC68U, and now the 87U, which is following the same trend (having the bugs worked out of the firmware over time). Tim's 87U review told it like it was at the time it was written: The hardware was intriguing, but the documentation was incomplete and spotty for many of the new feature sets (e.g., Adaptive QOS, etc), and there were as we know, a ton of bugs that needed to be worked out, including some with the 5ghz band that are still being worked out for many users. Oh, and when someone finally comes out with a 4x4 client (still waiting....tick-tock), maybe then we can all really know just how fast "fast" can be and how good the 87U can be.

But as for the "Tim-hate," I think it's enormously out of whack with what can objectively be found at this site, and if anything, I'd say that SNB gives almost more love and devotion to Asus than almost any other brand, and certainly more so than any other non-Asus-branded enthusiast site on the internet that I've seen. I mean, how else do you explain why there is not just a single forum devoted to Asus products, but three separate sub-forums exclusively about various Asus wireless products and firmware? Oh, yeah, could be because they are the only company that doesn't orphan users with planned obsolescence and instead continue to improve and support firmware many years after a product comes to market and long after newer and snazzier technology is released, and because there is a huge and avid user-base, including folks like Merlin, John and others who work on and develop independent firmware that make these products some of the best routers for SOHO wireless router enthusiasts.

Hey, we should all resolve to be better human beings and nicer people all the time. That certainly would include you as well, Wye.

I dont hate ASUS, i just hate that ASUS being a premium brand provides locked firmware regardless of region so i cant make use of all the wireless channels and transmission power unless i use unlocked firmware or change CFE.

If it werent for merlin firmware i wouldnt be using ASUS because merlin firmware keeps the stock performance while letting me use it as a mini linux server too while being less buggy.

I test my hardware with much more stress than what you guys normally do to the point where packets get dropped on LAN.

In my opinion Tim makes good articles. I was just teasing wye.

 

[jegesq]

SEM:

Understood. We cool.

 

[wye]

Hmm. I guess that's pretty far from your magical "5-10 Gigabytes/sec" SSDs.

Now that I look again, yeah, perhaps I was pretty aggressive in my OP indeed. I was pissed because I tried to gather some useful numbers from that AC87U review and I had to filter a lot of inaccurate numbers paired with the attitude.

Regarding my so called "magical" speed. The key is the "s" at the end of "SSDs". Multiple.

Initially I'll go with 2 x P3700 in software stripe, which is 5.6 read / 2.4 write (minus the raid overhead, ofc)
But I plan to expand to 5 x P3700, which is 14 read / 6 write. Which is way higher than my modest 5-10 "magical" numbers.

Everything sounds magical until you have the knowledge

 

[wye]

I would trust Tim Higgin's testing methods for measuring WAN to LAN throughput (described here for the N66U) far more than I would a test which measures only internet download speed using an online source like the Ookla engine, which can and often is wildly inaccurate and a very crude measure at best. It also isn't measuring WAN to LAN throughput. You could learn a thing or two also from Tim by reading some of the articles at SNB on the basics, including the differences between speed and throughput (See, e.g., http://www.smallnetbuilder.com/wireless/wireless-basics/32175-how-fast-can-your-wi-fi-go).

Regarding the term "speed" that you mention and your linked article is referring to, I never mentioned any negotiated link speed / link rate. Why would anyone? Why would you? Just throwing around random terms / semantic wars to confuse people?

Those are all throughput tests that I've done. Multiple times.
Sustained throughput, including hours of very practical torrent transfer, if you don't trust Ookla.

I actually did read the testing methodology used by Tim before my OP, and I believe the root cause is the blind trust in IxChariot paired with wrong settings used. Settings/software that was ok back then in 2010 or for weak routers. But not ok for 2014 with huge traffic. Tim, are you seriously using 100 kilobytes files to test gigabit traffic?

I suggest you run your own tests instead of religiously blindly trusting Tim's outdated methodology. And cut the whole "how dare you question our site owner/god" attitude. Anyone should be able to question anyone. Perhaps I'm right, perhaps I'm wrong. Don't feel insulted by skepticism, its the way science works. Instead, try to use logic, it will get you farther.

 

[wye]

PPPOE is a layer 2 protocol so the overhead isnt as much as you would think

I was referring to processing overhead, not envelope size.
I did ran tests for processing on PPPoE. Without hardware acceleration of PPPoE packet processing, the WAN to LAN throughput is about 180 Mbps. Pretty huge difference from 931 Mbps, so I guess its not something negligible.