Asus XT8 - Odd Behaviour on download from NAS attached to Node

[hoolieco]

My first time on the forums here but you all seem to have some pretty good insight on things. I've got 3 XT8s (One twin pack and one solo). Running the 23012 firmware. I've got one attached to my router as an AP and two nodes. PC on one node and NAS on another.

Now by default, the nodes seem to connect to the AP and when they sync up, I'm getting anything between 2.6Gbps and 2.8Gbps on one of them and 3.2 and 3.6Gbps so I'm pretty happy in this regard of wifi speeds. They seem really stable with pings across the network being consistent with very little if any dropped packets.

However, this behaviour seems to emerge - When I attempt to copy a file from my NAS (which has a throughput which capable of at least gigabit speeds), I'd expect based on the link speeds above to get circa 100MB/s off of a 1Gbps link. However it is nearly half that speed. The odd thing is, when go to the router webGUI and got to AIMesh, tell one of the nodes to speak to the other one instead of the AP, I get full gigabit (circa 100MB/s from the NAS). I can replicate it across a number of other firmwares.

I could leave the link to go to the other node but as a tradeoff from the speed, the link doesn't seem to be as stable (higher pings and some dropped packets) due to the distance between the two boxes.

So in summary, I have two nodes furthest apart from one another, using the AP as the middle-man, I get a speed penalty in favour of stable connections but connect to my other node using the web GUI and I get the full gigabit speeds with some instability. I've checked for other traffic across the nodes by disconnecting other devices from them

So my question is, any ideas why does this behaviour exist, if anyone knows, is there any fix?

Hope this is clear enough to answer but please ask for any more info.

 

[tgl]

Welcome to the forums, @hoolieco.

I may be misunderstanding your physical layout, but it sounds like the case you are measuring requires one node to transmit data from the NAS to the main router, which then retransmits to the other node, which feeds the PC that's collecting the data. Correct? If so, 50% throughput is the best you could get, because wifi is only half-duplex: the main router can't receive and transmit at the same time on the same channel. You could imagine setups where the two nodes use backhaul connections on different channels, but I don't think that's possible with ASUS' gear.

There are a bunch of other reasons why you'll never get full nominal throughput on wifi connections, and it might be that your real problem is something else. But in any case, you should try to avoid using wifi hops for gear that could perfectly well be wired. Why isn't your NAS connected to the main router instead of a node?

 

[hoolieco]

Thank you for your reply, your interpretation of my setup is correct. Purely down to the layout of my home.

I recall the half duplex issue with Wifi but I'd ask why do I see near perfect speed when connecting to the node that has the NAS attached, using the Preferred WiFi Uplink AP option in the Web GUI?

 

[tgl]

My point is that in that configuration, there's no retransmit involved: the node with the NAS is sending directly to the node with the PC. (There's still a lot of overhead, but not enough to drag the available throughput below the ~1Gbps limit of the end devices.)

 

[hoolieco]

Makes sense, thank you. To answer your question, the cost of a hardwire in my home isn't really an option just now but I do plan it in the future. The NAS is home built Unraid system with VM and Docker services which serve our main TV for a PC and Media Server. It's also got our main storage for all data. To reduce the costs involved in making another PC for the TV, it's a bit all in one. The node is used to service the TV, XBox and NAS.

Your explanation gives me some comfort on this that's a limitation more than anything. Thank you, again.

 

[degrub]

the highest NAS R speed will occur with long, contiguous reads. This is often quoted by the marketing dept.
Smaller files will read slower and are dependent on cache and buffer on the drive and NAS box. Also, it depends on the RAID level implemented. Look at the NASPT test results for more real world results for smaller files copied back and forth.

Use IPERF to test local LAN rates.

 

[hoolieco]

the highest NAS R speed will occur with long, contiguous reads. This is often quoted by the marketing dept.
Smaller files will read slower and are dependent on cache and buffer on the drive and NAS box. Also, it depends on the RAID level implemented. Look at the NASPT test results for more real world results for smaller files copied back and forth.

Use IPERF to test local LAN rates.

Thanks for the input, I was using one file of about 2 Gigabyte in size coming off of an SSD. So I'd expect and do see when hardwired, cira 100MBps. I'm comfortable with smaller files not giving the same throughput but I completely agree with your sentiment here.

The matter does seem to be retransmit due to half duplex. I'll just need to accept it for what it is.

 

[tgl]

The matter does seem to be retransmit due to half duplex. I'll just need to accept it for what it is.

More to the point: your effective wireless throughput is above 1Gbps until you throw in 50% overhead from retransmit, and then it isn't; indeed it seems you're getting only barely above 1Gbps before retransmit. If you're running at 160MHz channel width, your 4x4 backhaul radios have a theoretical limit of 4800Mbps, so we're led to the conclusion that you're only getting barely a quarter of that as effective throughput. That indeed seems more awful than one would expect. The usual rule of thumb is that if you get 65%-75% of the nominal Tx rate then you're doing fine; which you're not. There may be something to look into here, but with no data about your wireless configuration and environment we can't do more than guess.

 

[hoolieco]

More to the point: your effective wireless throughput is above 1Gbps until you throw in 50% overhead from retransmit, and then it isn't; indeed it seems you're getting only barely above 1Gbps before retransmit. If you're running at 160MHz channel width, your 4x4 backhaul radios have a theoretical limit of 4800Mbps, so we're led to the conclusion that you're only getting barely a quarter of that as effective throughput. That indeed seems more awful than one would expect. The usual rule of thumb is that if you get 65%-75% of the nominal Tx rate then you're doing fine; which you're not. There may be something to look into here, but with no data about your wireless configuration and environment we can't do more than guess.

What would be required to give more context? Logs?

 

[tgl]

What would be required to give more context? Logs?

The "Wireless Log" info from all three XT8s would be interesting to see. Also the Wireless/General configuration pages for the backhaul (5GHz-2) band, as well as Wireless/Professional if you've changed anything there.

One thing that that data isn't great at is showing how much environmental noise and interference you face. Which country do you live in (wifi regulations vary)? Are you in a densely populated neighborhood, or out in the sticks? Are you anywhere near an airport or weather radar?

 

[hoolieco]

The "Wireless Log" info from all three XT8s would be interesting to see. Also the Wireless/General configuration pages for the backhaul (5GHz-2) band, as well as Wireless/Professional if you've changed anything there.

One thing that that data isn't great at is showing how much environmental noise and interference you face. Which country do you live in (wifi regulations vary)? Are you in a densely populated neighborhood, or out in the sticks? Are you anywhere near an airport or weather radar?

Doing my own research and your contribution, I've identified that it's reverting to 80mhz, I'm in the UK and about 15 miles from Edinburgh Airport. The nearest weather radar is about 20-25 miles north of Edinburgh City. So these could all contribute, I believe.

Copy of wireless log:

SSID:
noise: -87 dBm Channel: 7
BSSID: 7C:10:C9:55:ED:10 Capability: ESS ShortSlot RRM
Supported Rates: [ 1(b) 2(b) 5.5(b) 6 9 11(b) 12 18 24 36 48 54 ]
HE Capable:
Chanspec: 2.4GHz channel 7 20MHz (0x1007)
Primary channel: 7
HT Capabilities: 40Mhz SGI20 SGI40
Supported HT MCS : 0-15
Supported VHT MCS:
NSS1 Tx: 0-11 Rx: 0-11
NSS2 Tx: 0-11 Rx: 0-11
Supported HE MCS:
80 Mhz:
NSS1 Tx: 0-11 Rx: 0-11
NSS2 Tx: 0-11 Rx: 0-11

Interference Level: Acceptable
Mode : AP Only

Stations List
----------------------------------------
idx MAC Associated Authorized RSSI PHY PSM SGI STBC MUBF NSS BW Tx rate Rx rate Connect Time
B0:4E:26:67:41:39 Yes Yes -65dBm n No Yes Yes No 1 20M 72.2M 72.2M 00:00:14
7C:10:C9:7A:6A:81 Yes Yes -58dBm ax No Yes Yes No 2 20M 229.4M 162.5M 00:00:37
7C:10:C9:7A:3A:C1 Yes Yes -41dBm ax No Yes Yes No 2 20M 206.5M 195M 00:00:37
4865:63:76:B8 Yes Yes -50dBm n No Yes No No 1 20M 72.2M 72.2M 00:00:43
48:60:5F:4CA:76 Yes Yes -64dBm n Yes Yes Yes No 1 20M 72.2M 6M 00:00:44
B8:27:EB:16:A3:C3 Yes Yes -68dBm n No Yes No No 1 20M 65M 57.8M 00:00:47

SSID:
noise: -87 dBm Channel: 60/80
BSSID: 7C:10:C9:55:ED:14 Capability: ESS RRM
Supported Rates: [ 6(b) 9 12 18 24(b) 36 48 54 ]
HE Capable:
Chanspec: 5GHz channel 58 80MHz (0xe23a)
Primary channel: 60
HT Capabilities: 40Mhz SGI20 SGI40
Supported HT MCS : 0-15
Supported VHT MCS:
NSS1 Tx: 0-11 Rx: 0-11
NSS2 Tx: 0-11 Rx: 0-11
Supported HE MCS:
80 Mhz:
NSS1 Tx: 0-11 Rx: 0-11
NSS2 Tx: 0-11 Rx: 0-11

Interference Level: Acceptable
Mode : AP Only

DFS status: state In-Service Monitoring(ISM) time elapsed 45750ms radar channel cleared by DFS channel 60/80 (0xE23A)

Channel Information
----------------------------------------
Channel 36 A Band
Channel 40 A Band
Channel 44 A Band
Channel 48 A Band
Channel 52 A Band, RADAR Sensitive
Channel 56 A Band, RADAR Sensitive
Channel 60 A Band, RADAR Sensitive
Channel 64 A Band, RADAR Sensitive

Stations List
----------------------------------------
idx MAC Associated Authorized RSSI PHY PSM SGI STBC MUBF NSS BW Tx rate Rx rate Connect Time
E2:2A:79:6C:10:3B Yes Yes -67dBm ax Yes Yes No No 2 80M 576.5M 6M 00:00:37
26:B0:97:24:61B Yes Yes -69dBm ac Yes Yes Yes No 2 80M 526.5M 6M 00:00:37
4A:A2:20:F9:33:C4 Yes Yes -60dBm ac Yes Yes Yes No 2 80M 780M 6M 00:00:38

SSID:
noise: -89 dBm Channel: 100/80
BSSID: 7C:10:C9:55:ED:18 Capability: ESS RRM
Supported Rates: [ 6(b) 9 12 18 24(b) 36 48 54 ]
HE Capable:
Chanspec: 5GHz channel 106 80MHz (0xe06a)
Primary channel: 100
HT Capabilities: 40Mhz SGI20 SGI40
Supported HT MCS : 0-31
Supported VHT MCS:
NSS1 Tx: 0-11 Rx: 0-11
NSS2 Tx: 0-11 Rx: 0-11
NSS3 Tx: 0-11 Rx: 0-11
NSS4 Tx: 0-11 Rx: 0-11
Supported HE MCS:
80 Mhz:
NSS1 Tx: 0-11 Rx: 0-11
NSS2 Tx: 0-11 Rx: 0-11
NSS3 Tx: 0-11 Rx: 0-11
NSS4 Tx: 0-11 Rx: 0-11
160 Mhz:
NSS1 Tx: 0-11 Rx: 0-11
NSS2 Tx: 0-11 Rx: 0-11
NSS3 Tx: 0-11 Rx: 0-11
NSS4 Tx: 0-11 Rx: 0-11

Interference Level: Acceptable
Mode : AP Only

DFS status: state In-Service Monitoring(ISM) time elapsed 42450ms radar channel cleared by DFS channel 100/80 (0xE06A)

Channel Information
----------------------------------------
Channel 100 A Band, RADAR Sensitive
Channel 104 A Band, RADAR Sensitive
Channel 108 A Band, RADAR Sensitive
Channel 112 A Band, RADAR Sensitive
Channel 116 A Band, RADAR Sensitive, Passive
Channel 120 A Band, RADAR Sensitive, Passive
Channel 124 A Band, RADAR Sensitive, Passive
Channel 128 A Band, RADAR Sensitive, Passive
Channel 132 A Band, RADAR Sensitive, Passive
Channel 136 A Band, RADAR Sensitive, Passive
Channel 140 A Band, RADAR Sensitive

Stations List
----------------------------------------
idx MAC Associated Authorized RSSI PHY PSM SGI STBC MUBF NSS BW Tx rate Rx rate Connect Time
7C:10:C9:7A:6A:88 Yes Yes -54dBm ax No Yes Yes Yes 4 80M 1441.2M 1441.2M 00:00:39
7C:10:C9:7A:3A:C8 Yes Yes -42dBm ax No Yes Yes Yes 4 80M 2041.7M 2161.8M 00:00:40

 

[tgl]

Doing my own research and your contribution, I've identified that it's reverting to 80mhz, I'm in the UK and about 15 miles from Edinburgh Airport. The nearest weather radar is about 20-25 miles north of Edinburgh City. So these could all contribute, I believe.

Yeah. The wireless log clearly shows that you're running at 80MHz not 160MHz channel width, which pretty much accounts for the performance discrepancy from my back-of-the-envelope calculation. But 4x4 radios in 80MHz channels should have best-case Tx rates of 2400MHz, so it's not quite clear how you saw Tx rates above that originally. If you have the things set for variable channel width then it's possible that they run at 160MHz until they see a radar pulse and then drop to 80MHz for awhile, which is about the best behavior you can hope for under UK regulations.