Community discussions

MikroTik App
 
AKSN74
just joined
Topic Author
Posts: 11
Joined: Sun Apr 26, 2015 5:06 pm
Location: Taichung, Taiwan.

Wireless performance on hAP ac (non lite)

Fri Jun 24, 2016 1:45 pm

Hello everyone,
Few days ago I got hAP ac and test the speed about wireless.

Environment:
1 PC (Windows 10) connect on hAP ac with Gigabit LAN
1 Macbook Pro 13" (early 2015) connect on hAP ac with 802.11ac wireless
Distance between hAP ac and Macbook Pro: 2m (got little block with desk.)
Noise: No 5G noise.

Then open the file sharing on Mac, test download files from Mac and upload files to Mac with PC.
Here is my result

Download from Mac:
Image

Upload to Mac:
Image

The mac address F4:5C in registration is my Macbook Pro.


And here is my 5G wireless settings:
Image 

I had two questions:

1. When I downloading files from Macbook Pro, the average speed is close to 60MB/s, highest could be about 63~64MB/s
But TX will jump between 2S and 3S.
I got same situation when uploading, higher speed rate, but upload speed is not as good as downloading (about 38~42MB/s)
(I guess is Mac OS X Samba settings)
Is that normal for TX chains jumping?

2. When I downloading files from Macbook Pro, high speed, but also high CPU usage (about 90~95%).
And I sure that whole wireless transfer running with FastPath.
How can I reduce CPU usage?

Many thanks,
AK.
 
mpreissner
Member
Member
Posts: 357
Joined: Tue Mar 11, 2014 11:16 pm
Location: Columbia, MD

Re: Wireless performance on hAP ac (non lite)

Fri Jun 24, 2016 4:59 pm

What you have to realize is that the wireless interfaces have to be bridged to the wired interfaces.  The bridge is implemented in software, so it consumes a lot of CPU.  Any high-speed AC wireless transfers will result in a good amount of CPU usage.  Even if you assigned a completely separate subnet to the wireless interface, you'd be routing the traffic versus bridging it, which would still consume a fair amount of CPU.

I can't speak for MikroTik's wireless products, but most wireless products are optimized for high download rates, as the majority of client traffic is downstream in direction.  Also, Apple's implementation of Samba for SMB/CIFS data transfer is not quite as optimized as their AFP protocol data transfers, so don't expect the best results using smb.
 
AKSN74
just joined
Topic Author
Posts: 11
Joined: Sun Apr 26, 2015 5:06 pm
Location: Taichung, Taiwan.

Re: Wireless performance on hAP ac (non lite)

Fri Jun 24, 2016 6:40 pm

What you have to realize is that the wireless interfaces have to be bridged to the wired interfaces.  The bridge is implemented in software, so it consumes a lot of CPU.  Any high-speed AC wireless transfers will result in a good amount of CPU usage.  Even if you assigned a completely separate subnet to the wireless interface, you'd be routing the traffic versus bridging it, which would still consume a fair amount of CPU.

I can't speak for MikroTik's wireless products, but most wireless products are optimized for high download rates, as the majority of client traffic is downstream in direction.  Also, Apple's implementation of Samba for SMB/CIFS data transfer is not quite as optimized as their AFP protocol data transfers, so don't expect the best results using smb.
Yeah, you're right, seems like Apple didn't do a good work at samba.

In other way, as the result, it may not be a good performance if the internet is using FTTH (which using PPPoE) and high download speed using wireless
For my RB2011UiAS, it can only get about 22x Mbits/s download speed when using Ethernet to test with no firewall rule set.

I guess hAP ac can get 300M download throughput in Ethernet, but I'm not sure it can in wireless
(Our ISPs sucks, you can't connect to their network with your own modem, you have to use modem that ISP give and it's already about 5~10 years ago product)

And not talk about Mikrotik wireless, is other wireless product using the same design? Because I remember other wireless router still need to connect switch chip by CPU.
 
mpreissner
Member
Member
Posts: 357
Joined: Tue Mar 11, 2014 11:16 pm
Location: Columbia, MD

Re: Wireless performance on hAP ac (non lite)

Sat Jun 25, 2016 3:08 pm

So the one thing to remember is that you will ALWAYS get better performance over a wire than using wireless.  I try to wire everything that I can and minimize my use of wireless, not just for security reasons, but also just for plain old performance.  I can't speak to how good MT's wireless products are...I don't personally use them.  I use Ubiquiti's UniFi wireless products, as do many people on this forum (but there are plenty of MT users as well!).  No matter whose wireless tech you use, I always find it better to use dedicated APs instead of using an all-in-one product like the hAP, or similar stuff from Linksys, Cisco, NetGear, D-Link, etc.  This type of setup ends up costing more (not much if you do it right), but it much more flexible, and a lot easier to diagnose performance issues because the wireless is totally separate from the wired network.  If the wireless bridge doesn't have to share CPU with any other functions (wired bridges, routing, VPN, etc.) then the performance is usually much better, as long as you're using a quality product.
 
AKSN74
just joined
Topic Author
Posts: 11
Joined: Sun Apr 26, 2015 5:06 pm
Location: Taichung, Taiwan.

Re: Wireless performance on hAP ac (non lite)

Sat Jun 25, 2016 5:23 pm

So the one thing to remember is that you will ALWAYS get better performance over a wire than using wireless.  I try to wire everything that I can and minimize my use of wireless, not just for security reasons, but also just for plain old performance.  I can't speak to how good MT's wireless products are...I don't personally use them.  I use Ubiquiti's UniFi wireless products, as do many people on this forum (but there are plenty of MT users as well!).  No matter whose wireless tech you use, I always find it better to use dedicated APs instead of using an all-in-one product like the hAP, or similar stuff from Linksys, Cisco, NetGear, D-Link, etc.  This type of setup ends up costing more (not much if you do it right), but it much more flexible, and a lot easier to diagnose performance issues because the wireless is totally separate from the wired network.  If the wireless bridge doesn't have to share CPU with any other functions (wired bridges, routing, VPN, etc.) then the performance is usually much better, as long as you're using a quality product.
So it seem like this is general architecture, so they all using CPU resource to transfer wireless data in wireless routers.
And yes, dedicated APs is much flexible and stable, especially many wireless users or high speed transfer.
Maybe we're a same, I'm using UniFi AP AC, too ! (first generate, which is square modeling) 
Set it as AP, and connect with RB2011UiAS.

Anyway, thanks your explain, I learned a lot !
(And sorry about my bad English :) )
 
sstyle
Frequent Visitor
Frequent Visitor
Posts: 50
Joined: Thu Jul 14, 2016 11:53 pm

Re: Wireless performance on hAP ac (non lite)

Tue Jul 19, 2016 6:54 pm

Did you find a solution? I am in same boat with you. Macbook pro 2015 and hap ac. maximum transfer speed is about 550Mbit\sec.
 
paulct
Member
Member
Posts: 336
Joined: Fri Jul 12, 2013 5:38 pm

Re: Wireless performance on hAP ac (non lite)

Wed Jul 20, 2016 9:51 am

I think 550Mbps over the wireless is pretty darn good. Remember the wireless specifications are a theoretical airspeed rate. Much like one cannot achieve 300Mbps over wireless N.
 
sstyle
Frequent Visitor
Frequent Visitor
Posts: 50
Joined: Thu Jul 14, 2016 11:53 pm

Re: Wireless performance on hAP ac (non lite)

Wed Jul 20, 2016 9:53 am

I think 550Mbps over the wireless is pretty darn good. Remember the wireless specifications are a theoretical airspeed rate. Much like one cannot achieve 300Mbps over wireless N.
It is very good but it can be better. Look this https://forum.openwrt.org/viewtopic.php?id=53703 I believe that mikrotik can do the same with correct settings
 
paulct
Member
Member
Posts: 336
Joined: Fri Jul 12, 2013 5:38 pm

Re: Wireless performance on hAP ac (non lite)

Wed Jul 20, 2016 10:40 am

I think 550Mbps over the wireless is pretty darn good. Remember the wireless specifications are a theoretical airspeed rate. Much like one cannot achieve 300Mbps over wireless N.
It is very good but it can be better. Look this https://forum.openwrt.org/viewtopic.php?id=53703 I believe that mikrotik can do the same with correct settings
I have an Archer C7, but no 3x3 ac adapter - should get one soon to test. However most other people using the C7 cannot seem to get the speeds that the first poster did. I would be stoked to get 800mbps :) but I do not sit 0.5m away from the router. I suppose the C7 would be better at longer distances than the hap ac. Would be interesting to see a UDP iperf test on the Hap AC vs a TCP test with CPU usage on both.
 
Zorro
Long time Member
Long time Member
Posts: 675
Joined: Wed Apr 16, 2014 2:43 pm

Re: Wireless performance on hAP ac (non lite)

Wed Jul 20, 2016 7:19 pm

I think 550Mbps over the wireless is pretty darn good. Remember the wireless specifications are a theoretical airspeed rate. Much like one cannot achieve 300Mbps over wireless N.
its very depend on radio spectrum condition in link location and whole nature of connection(eg how narrowbeam it, how powerful and sensitive, how Properly configured and frequently maintained and etc).
and speeds was conviniently keep skyrocketing as well as complexity of randos(and antennas array, sadly :(
"expected to be released/finalised" 802.11ax for example - would had up to 1200 Mbps peak even in "baseline" profile without various tricks/shortcuts. and tripple-band(60Ghz was 3rd. sib-mm bands radios kinda stay "too expensive", despite sexy ISM bands lying out there too).
 
paulct
Member
Member
Posts: 336
Joined: Fri Jul 12, 2013 5:38 pm

Re: Wireless performance on hAP ac (non lite)

Thu Jul 21, 2016 9:23 am

I think 550Mbps over the wireless is pretty darn good. Remember the wireless specifications are a theoretical airspeed rate. Much like one cannot achieve 300Mbps over wireless N.
its very depend on radio spectrum condition in link location and whole nature of connection(eg how narrowbeam it, how powerful and sensitive, how Properly configured and frequently maintained and etc).
and speeds was conviniently keep skyrocketing as well as complexity of randos(and antennas array, sadly :(
"expected to be released/finalised" 802.11ax for example - would had up to 1200 Mbps peak even in "baseline" profile without various tricks/shortcuts. and tripple-band(60Ghz was 3rd. sib-mm bands radios kinda stay "too expensive", despite sexy ISM bands lying out there too).
Yup I was looking at the 802.11ad 60Ghz from TP-link, but with no compatible devices on my home network (yet) I will wait a couple years to see if it catches on and more devices to support it. I like the idea of it in my lounge where my NAS, AP, TV, android box sits i.e close proximity for 60ghz.
Anyway, enough off topic ;)
 
Zorro
Long time Member
Long time Member
Posts: 675
Joined: Wed Apr 16, 2014 2:43 pm

Re: Wireless performance on hAP ac (non lite)

Thu Jul 21, 2016 10:04 am

to OP: try play with WMM and multicast helper(try it in "full").
also try different ANI modes and try enable (both)frame protection.
also temporally disable "multicast buffering" and ensuare that "polling" was disabled too.
do "spectral scan" and see where is less noise in you location and less output in devices in proximity.
and last but not less important: with such Small distance - you devices may simply "scream too loud" and RX on both sides start perform sub-optimal. try reduce TX power by several steps.
 
sstyle
Frequent Visitor
Frequent Visitor
Posts: 50
Joined: Thu Jul 14, 2016 11:53 pm

Re: Wireless performance on hAP ac (non lite)

Thu Jul 21, 2016 11:00 am

also try different ANI modes and try enable (both)frame protection.
and last but not less important: with such Small distance - you devices may simply "scream too loud" and RX on both sides start perform sub-optimal. try reduce TX power by several steps.
Thanks, but I can not find ANI and frame protection ( Can you give an exact names of this settings?
About distance: I see a 1300mbit connection speed very rarely and when my macbook very closer to router. Sure I try to decrease TX-power. It is possible if increase antenna gain. But it drops my real transfer speed. Best result was with 23dBm and 4m distance.
 
Zorro
Long time Member
Long time Member
Posts: 675
Joined: Wed Apr 16, 2014 2:43 pm

Re: Wireless performance on hAP ac (non lite)

Thu Jul 21, 2016 11:24 am

also try different ANI modes and try enable (both)frame protection.
and last but not less important: with such Small distance - you devices may simply "scream too loud" and RX on both sides start perform sub-optimal. try reduce TX power by several steps.
Thanks, but I can not find ANI and frame protection ( Can you give an exact names of this settings?
About distance: I see a 1300mbit connection speed very rarely and when my macbook very closer to router. Sure I try to decrease TX-power. It is possible if increase antenna gain. But it drops my real transfer speed. Best result was with 23dBm and 4m distance.
okay
for example ANI http://gregsowell.com/wp-content/upload ... 08/ani.png
generally you may notice rest, including "Frame protection support" described in official ROS Wiki
http://wiki.mikrotik.com/wiki/Manual:In ... S.2FCTS.29
http://wiki.mikrotik.com/wiki/Manual:Interface/Wireless (its not updated for wireless-rep package performance and feats, yet).
and no, you don't get it -try decrease TX power WITHOUT touching "antenna gain".
 
sstyle
Frequent Visitor
Frequent Visitor
Posts: 50
Joined: Thu Jul 14, 2016 11:53 pm

Re: Wireless performance on hAP ac (non lite)

Thu Jul 21, 2016 12:13 pm

oh, I see, did it. Did not help to increase a speed. 
There is no settings for decrease tx-power another way. I asked about it here - http://forum.mikrotik.com/viewtopic.php?f=7&t=110316

Who is online

Users browsing this forum: jaclaz, Majestic-12 [Bot] and 40 guests