Range Extenders ... Is "Conventional Wisdom" Correct?

[Klueless]

Edit: Ok, I'm an idiot. If I could delete this entire thread I would. Until then it remains as a monument to my stupidity. My benchmarks misled me. My throughput never got halved. In fact throughput was getting better. To some extent it should. Otherwise why put in a range extender?

But I was missing the whole point. I send something to the range extender. Then I can't. I can't send something to the range extender because it is using the radio to send my stuff to the router. When it's done then I can send something. Maybe. Maybe not. The guy at the next desk might be sending something by then.

My data was showing me the benefit of putting in a range extender, it was showing me that my life is better than it used to be. It was showing my cup as half full.

But my data was not showing me what my life could be ... as compared to a wired AP. It was not showing me my cup was really half empty.

We've all seen it, "Use a wired AP. When you use a range extender your times will be halved." At first blush it makes sense. You send something and it takes x time, send it twice - 2x. So you send to a range extender, it takes x. Then the range extender receives and sends it again thus the total trip must take about 2x or, said another way, the speed is halved.

It all makes sense except it doesn't; my benchmarks were showing only a 25% hit.

I placed my laptop near a window facing the bldg. next door where our router is. I connect using N at 2.4Ghz.

• I get 28 Mbps throughput.

So if I connect to the 2.4Ghz B/G/N range extender located by the same window. I should see about 14 Mbps.

• Wrong. I get 21 Mbps. Instead of a 50% hit I only get a 25% hit. Most pleasant surprise!

Now (long story) I've got a new range extender that runs at 2.4Ghz and 5Ghz. I use the 5Ghz as an air-wire to the main bldg. and use the 2.4 to feed the users in our bldg.

I bench with my laptop near the window and connect directly to the main bldg. at 5Ghz.

• I get 30.5 Mbps throughput.

Hmm ... half would be 15 but 25% would be about 23 Mbps. (Lotta work for that little bit of improvement but hopefully it'll reduce all the variability we've been experiencing.) So, holding my breathe, I connect to the new range extender at 2.4Ghz)

• And ... the big momment ==> 50 Mbps! (Yes, 50. Instead of a decrease I get a significant increase in performance. Is that crazy or what?)

Even better everything seems solid; packet loss is gone, user perception is better, Internet speed tests are solid.

My thoughts (all of them probably wrong):

• Maybe the extender doesn't wait for the whole packet to arrive before it starts sending? Instead of "store and forward" does it do a "fast switch"? <edit> Pete politely says, "no, you have to wait for the radio to clear"
• The formula for speed is not 2x (x + x). It's actually x' + y.
  • Changing the speed of x (now x') or y will improve performance:
    • y (client to range extender) should be faster than (the original) x. Wasn't that why we put in the range extender to start with?
    • Slight re-positioning of range extender (especially at 5Ghz) can make a (big) difference.
      • In my case instead of placing it on the desk near the window (where I ran my benchmarks) I actually Velcro'd it to the window.
    • There was an article on this site that running N on an AC router runs faster than running N on an N router. While I've only an N router my range extender is AC. Am I getting a similar (somewhat inexplicable) boost?
    • Using 5Ghz as a back-haul (only) increased speed while reducing the collision domain (also increasing speed while reducing variability). If I wasn't so lazy I'd actually consider setting the 5Ghz on the router to connect to the range extender only (further reducing the collision domain).
      • <edit> SEM suggests no clients on the radio you use for the backhaul. I did that for the bldg. 2 (Y) side but it's not a luxury I can easily afford on the main bldg. (X) side. Products that lean towards SEM's thought process include:
        • Netgear's Fastlane
        • Linksys' Cross Band
    • Wired APs have a similar send twice phenomenon but they compensate by drastically altering x' (the back-haul)
      • 100 to 1,000 Mbps (depending on cable and distance)
      • Fixed rather than varying link rates
      • Dedicated line, not shared, no collision domain.
      • Full duplex (send and receive don't share bandwidth)

Now I'm not out to be right, I'm just out to learn. Any/all comments are most certainly welcome.

 

[pete y testing]

i think you mis understand what is meant when they say the repeater 1/2's time

how a repeater works is it receives a signal from the primary source and the receptions has to stop to then re transmit , thus the duty cycle time is doubled eg receive on transmit off / receive off transmit on and the other way around for the client device , look at it as hops , if a signal goes direct from the main wifi to the client its one hop in each direction , whereas if a repeater is used its 2 hops in each direction


this hit will be see both as latency increase and throughput decrease but it will be seen from the read and the write times and the time it takes to send and receive / read and write

you increase is mainly due to using a new better repeater that is connecting at a far better sync rate thus better throughput

Now I'm not out to be right,

not quite

Wired APs have the same send twice phenomenon but they compensate by drastically altering the speed of x' (the back-haul)

not sure what you are saying here but wireless is half duplex anyway and adding a repeater makes the 1/2 duplex half again dont forget wifi is a 2 way conversation

 

[sfx2000]

We've all seen it, "Use a wired AP. When you use a range extender your times will be halved." At first blush it makes sense. You send something and it takes x time, send it twice - 2x. So you send to a range extender, it takes x. Then the range extender receives and sends it again thus the total trip must take about 2x or, said another way, the speed is halved.

It all makes sense except it doesn't; my benchmarks were showing only a 25% hit.

It doesn't gut bandwidth, but an extender/repeater will use up twice as much time...

Here's the kicker though - if one has a client that is out on the edge, and taking up a huge amount of time at low bandwidth, putting in a repeater can help out, because each leg (from client to repeater, and from repeater to AP) can be much faster...

Which may explain why you only see a 25 percent hit, instead of 50 percent...

WiFi is a shared medium - and time is the fire in which it burns - the idea is to get onto the medium, send/receive the bits, and clear the channel for someone else...

 

[Klueless]

i think you mis understand what is meant when they say the repeater 1/2's time

You are correct. I was mislead by the context of articles I read as they implied my throughput would be halved.

Thank you.

 

[sfx2000]

You are correct. I was mislead by the context of articles I read as they implied my throughput would be halved.

In best conditions - thru-put is halved with a repeater...

 

[Klueless]

It doesn't gut bandwidth, but an extender/repeater will use up twice as much time...

Here's the kicker though - if one has a client that is out on the edge, and taking up a huge amount of time at low bandwidth, putting in a repeater can help out, because each leg (from client to repeater, and from repeater to AP) can be much faster...

Which may explain why you only see a 25 percent hit, instead of 50 percent...

Makes sense. (Where were you before I posted this stupid thread? Where's the delete button? : -)

I think you also make a case for using the same SSID/passwords on APs such that as you reach an edge you move to a new area rather than pulling everyone down by staying on the edge? (Wouldn't work with a repeater though, would it?)

Hmm ... could that mean if all my neighbors all used the same SSID we could walk down the entire street without losing connection?

 

[sfx2000]

I think you also make a case for using the same SSID/passwords on APs such that as you reach an edge you move to a new area rather than pulling everyone down by staying on the edge? (Wouldn't work with a repeater though, would it?)

Depends on the client - but many will make the jump just fine these days... even thru a repeater - might be a small gap to resync keys, but generally, they'll complete the jump to hyperspace and land accordingly...

(good discussion, so don't hit the delete key)

 

[Klueless]

Depends on the client - but many will make the jump just fine these days... even thru a repeater - might be a small gap to resync keys, but generally, they'll complete the jump to hyperspace and land accordingly

My impression was clients will jump from SSID to SSID if they are open or from SSID to SSID if you've already configured the password but if all the SSIDs are the same the jump is quicker, more likely and with less overhead?

 

[Klueless]

In best conditions - thru-put is halved with a repeater...

In a best condition I had run the same bench one office away from the router (both in the main building) and got 65.6 Mbps.

In bldg. 2, connecting through the range extender, I got 50 Mbps. That's about a 24% hit, not 50%. I don't understand, I may never understand, but I am delighted!

 

[sfx2000]

My impression was clients will jump from SSID to SSID if they are open or from SSID to SSID if you've already configured the password but if all the SSIDs are the same the jump is quicker, more likely and with less overhead?

Different SSID's - client will try it's best to stay on the current one until it's lost...

Same SSID, depends on the client, but generally many will hop onto the strongest source...

 

[sfx2000]

In a best condition I had run the same bench one office away from the router (both in the main building) and got 65.6 Mbps.

Like the Book of Faces (ant-ispam), things are complicated...

Move the repeater left or right if the AP is on one side, and the client is on the other...

Repeaters generally have the same Tx power as AP's, clients generally have about half of that in many/most cases...

 

[stevech]

this is true: If a client device's signal to/from the WiFi router is quite weak, the 802.11 bit rate will "slow" a lot due to the poor signal to noise ratio (SNR). So the air time use increased due to the slowed bit rate. To that one client device, the weak signal slows d0wnloads and responses. But a handheld doesn't usually doesn't do heavy traffic with lots of data.

That weak signal's increased of air time hurts other clients with good signals and higher data rates.

It's all statistics and so on. What clients' demands are, where they are, etc. At the air-time-use level, matters not what SSID is using the air time - 802.11 is a listen-before-transmitting wireless method; half-duplex - cannot transmit and receive simultaneously (unlike most cellular).

So the general practice in shared-use WiFi/802.11 is try to design layout for great signal strength BOTH DIRECTIONS to minimize air-time use, reduce retransmissions for error correction, etc.
The repeater-range extenders with one radio, say 2.4GHz in and out, have to receive then retransmit each packet. That doubles air-time use since only one channel is used. Some few use a 2.4GHz channel for client access and 5GHz for connection to the WiFi router. This works if the 5GHz link is viable given where the repeater is placed. It's a challenge to find the best location for both the link to the WiFi router and creating the new 2.4GHz coverage footprint for the repeater. And that footprint on 2.4GHz may overlap the WiFi router's 2.4GHz coverage footprint - leading to more airtime use competition. And on and on.

Remember: the coverage footprint for the from-router transmissions is usually much larger that the footprint of the from-client - because clients tend to have less power and an inferior antenna. Two fried-egg shaped patterns of greatly different sizes. Not simple circles, due to walls, floors, furniture, body-blockage, etc.

 

[Klueless]

That weak signal's increased of air time hurts other clients with good signals and higher data rates.

You and SFX just turned on a light for me. I hadn't fully considered the dynamics of all this. I had been me-centric in thinking of them as my legs (which I refer to as X and Y). The reality is they are our legs and a leg is only as good as the weakest client at that moment.

So my changes made some other client's life much better such that he made my life much better which could explain why my benchmarks were much better than what I would have expected?

So, if I were to re-calibrate and re-bench everything I should see results more consistent with what everyone has been telling me?

 

[System Error Message]

It really depends on the repeater and how it connects. In order to prevent halving of throughput you would be using different channels and dedicating radios for bridging.

But even in the worse setup of using an extender you would be expecting 50% of the maximum theoratical throughput. That means if your maximum theoriatical is 300Mb/s than you would be expecting a maximum of 150Mb/s. The keyword is theoratical.

Honestly if you need to use wireless range extender please do it properly in which you would be using directional antennas, dedicating 2 radios (1 on each end) for bridging and using another radio to distribute wifi on a different channel. I cannot stress how unreliable wireless range extenders are if using the same radio to distribute wifi also for bridging and using omni antennas.

 

[Klueless]

So, if I were to re-calibrate and re-bench everything I should see results more consistent with what everyone has been telling me?

Ok. Here's a fresh set of benches, the original setup is the topmost and the bottom set is with the new range extender. (You may have to login and/or click to see image full size.)

And ... I'm not seeing anything that screams 50% hit? I assume I should base my calculation on the 38 Mbps I got from the window in the White House? (The same window my range extender is Velcro'd to.) That should give me a roughly right "best case" to compare against the range extender to see how much time it adds? So half of 38 should be 19 rather than the 50 Mbps I'm seeing?

 

[Klueless]

Thank you very much for taking the time to comment.

It really depends on the repeater and how it connects. In order to prevent halving of throughput you would be using different channels and dedicating radios for bridging.

Doesn't seem to be an option with the consumer grade products I'm using.

But even in the worse setup of using an extender you would be expecting 50% of the maximum theoratical throughput. That means if your maximum theoriatical is 300Mb/s than you would be expecting a maximum of 150Mb/s. The keyword is theoratical.

I believe my router is N 450 @ 5Ghz and N 450 again @ 2.4Ghz. I'm not seeing anywhere near that anywhere. (See my benchmarks in previous post.)

Honestly if you need to use wireless range extender please do it properly in which you would be using directional antennas, dedicating 2 radios (1 on each end) for bridging and using another radio to distribute wifi on a different channel. I cannot stress how unreliable wireless range extenders are if using the same radio to distribute wifi also for bridging and using omni antennas.

Brings back a few unhappy memories. When I first heard they were talking about expanding into the bldg. next door I had contractors come in to propose how to connect the two bldgs together. One was an Ethernet over radio solution (which would be somewhat similar to what you were suggesting) and the others were hard wire. Anyway, everything I brought to the table was turned down so ... here we are.

 

[stevech]

The conventional and widely used method, if cat5 or cat6 or fiber just CANNOT be used between buildings, is a short-haul 5GHz bridge pair. Engenius is a good one.
Or a good (cisco aironet) 5GHz bridge pair with a small 25dBi dish for 10 miles or so, on rooftops.

 

[System Error Message]

Thank you very much for taking the time to comment.

Doesn't seem to be an option with the consumer grade products I'm using.

I believe my router is N 450 @ 5Ghz and N 450 again @ 2.4Ghz. I'm not seeing anywhere near that anywhere. (See my benchmarks in previous post.)

Brings back a few unhappy memories. When I first heard they were talking about expanding into the bldg. next door I had contractors come in to propose how to connect the two bldgs together. One was an Ethernet over radio solution (which would be somewhat similar to what you were suggesting) and the others were hard wire. Anyway, everything I brought to the table was turned down so ... here we are.

In the first quote i meant theoratical. your benchmark was still lower than 50% theoratical.

 

[thiggins]

Apologies if you have already covered these points.

First, how are you measuring throughput?
What is the make/model of the repeater?
What is the make/model of the wireless adapter you are using for your benchmarks?

The repeater has to manage two links, one to the base AP, the other to the far-end STA (client). These can be at two entirely different rates, even with one radio.

The 50% rule assumes both links are the same so that airtime will be the same for each frame that is moved through the repeater. If one of the links is faster, it will take less airtime and the throughput reduction will be less than 50%.

Today's wireless standards use many different compression, modulation and airtime management techniques. You'd have to dig down to at least the next level of looking at transmit and receive link rates to really figure out what's going on.

 

[Klueless]

Apologies if you have already covered these points.

1. First, how are you measuring throughput?
2. What is the make/model of the repeater?
3. What is the make/model of the wireless adapter you are using for your benchmarks?

1. Netstress from my laptop to range extender to router (Asus RT-N66R) through Ethernet to a Desktop PC
2. Netgear EX6150 AC1200
3. Intel(R) Centrino(R) Advanced-N 6205

The repeater has to manage two links, one to the base AP, the other to the far-end STA (client). These can be at two entirely different rates, even with one radio.

The 50% rule assumes both links are the same so that airtime will be the same for each frame that is moved through the repeater. If one of the links is faster, it will take less airtime and the throughput reduction will be less than 50%.

Today's wireless standards use many different compression, modulation and airtime management techniques. You'd have to dig down to at least the next level of looking at transmit and receive link rates to really figure out what's going on.

I think you're saying what I was trying to postulate. The formula for time is not x + x (2x) but rather x + y.

• x is the original link (which should now be faster since you've moved the anchors off of it and onto the range extender)?
• y is the new link (connecting clients to range extender). It should be faster than x time or why even put it in?

Thus x + y should be less than 2x? Less than a 50% hit should be the norm?

My most pleasant conundrum is my x + y (throughput wise) is even better than just plain old x. (See benchmarks a few messages back.)

I take a bench from bldg 2 to bldg 1. 38 Mbps. My range extender is in the same window from where I took my bench. I take a bench through the range extender. Conventional wisdom suggests 19 Mbps. You(?) and I are thinking somewhere between 19 and 38 (maybe 28-ish?) Bam. I now get 50 Mbps!