Controllerless Networks

With an iAP-215 and 802.11ac NIC having the fastest possible a-VHT-80sgi-3ss (1300Mbits/sec) connection, the maximum performance I'm seeing is 341Mbits/sec with iperf 2 and 464Mbits/sec ftp transferring a 2.3GB compressed file.

Should I stop bludgening my head expecting to see something near 1300Mbits/sec performance because real world speed of 802.11ac is actually not that fast?

iperf -c server -t 10 -P10 -w 64k = 341Mbits/sec

I know the source and target are more than fast enough: Using a 1GB wired Ethernet NIC connection, the same tests fills the pipe at 960Mbits/sec, and on a 10GB LAN connection, it blows through at 3500Mbits/sec.

The iAP-215 with 802.11ac is rated at 1300Mbits/sec (which is 300Mbits faster than 1Gbit Ethernet) (doesn't that beg the question, shouldn't the AP have a 10Gbit Ethernet port?)

The client uses an ASUS AC1900 USB-AC68 USB 3.0 WiFi adapter that supports 1300Mbits/sec at 3 Spacial Streams (which is the only way to get a 1300Mbits connection).

Client adapter is five feet from iAP-215, no obstructions between. AP and client are at desk height and oriented toward each other.  iAP-215 is powered with 802.11at POE.

Can anyone else check their results, please?

BTW, the trick to get a wireless adapter to work with Windows Server 2012 R2 is to install the latest manufacturer's driver and then add the WS2012 R2 WLAN Feature (Server Manager, Manage, Add Roles and Features) which will then reveal a "Connect/Disconnect" menu option when right-clicking on the Wireless Adapter in Network Connections.  Subsequently (perhaps after a Windows restart), the traditional Wireless icon will appear in the System Tray.

show ap association

The phy column shows client's operational capabilities for current association

Flags: A: Active, B: Band Steerable, H: Hotspot(802.11u) client, K: 802.11K client, M: Mu beam formee, R: 802.11R client, W: WMM client, w: 802.11w client V: 802.11v BSS trans capable

PHY Details: HT   : High throughput;      20: 20MHz;  40: 40MHz; t: turbo-rates (256-QAM)

             VHT  : Very High throughput; 80: 80MHz; 160: 160MHz; 80p80: 80MHz + 80MHz

             <n>ss: <n> spatial streams

Association Table

-----------------

Name               bssid              mac                auth  assoc  aid  l-int  essid   vlan-id  tunnel-id  phy              assoc. time  num assoc  Flags  DataReady

----               -----              ---                ----  -----  ---  -----  -----   -------  ---------  ---              -----------  ---------  -----  ---------

04:bd:88:xx:xx:xx  04:bd:88:xx:xx:xx  34:97:f6:xx:xx:xx  y     y      2    1      ATest1  3333     0x0        a-VHT-80sgi-3ss  19m:28s      1          WK     Yes (Implicit)

Num Clients:1

*** Note: a-VHT-80sgi-3ss indicates the fastest connection possible for the iAP-215.

*** Note that ftp throughput with a 2 Spacial Stream (a-VHT-80sgi-2ss) connection is about 50MBytes/sec.  Windows Network Connection Adapter Status displayed speed rate for a a-VHT-80sgi-2ss connection is typically 866Mbits.

show clients

Client List

-----------

Name                     IP Address     MAC Address        OS       ESSID   Access Point       Channel  Type  Role    Signal    Speed (mbps)

----                     ----------     -----------        --       -----   ------------       -------  ----  ----    ------    ------------

WS2012-R2          172.31.98.133  34:97:f6:xx:xx:xx           ATest1  04:bd:88:xx:xx:xx  36E      AC    ATest1  53(good)  1300(good)

Number of Clients   :1

TIA!

What channel are you using for 5GHz?

FYI: 1,3Gbps is a theoretical speed. I doubt there is currently a device capable of maximizing throughput on AC. You'd also need the AP to be in either 80MHz or 160MHz. Which is not that common, especially in business environments were channel bonding is disabled as to maximize the amount of channels that can be used. However, some models are provisioned with multiple ports and allow aggregates to form 2Gb uplinks (just in case they eventually reach that throughput). 

---

@nuit wrote:

Using an iAP-215, at best, I'm experiencing iperf and ftp compressed file transfers of 50MBytes/sec using a client 802.11ac NIC. Should I stop bludgening my head expecting to see greater performance because this is simply the actual real world speed of 802.11ac?

 

I know the source is more than fast enough: With a 1GB LAN connection, same tests fill the pipe at nearly 120MBytes/sec.  With a 10GB LAN connection, 3500MBytes/sec.

 

The iAP-215 with 802.11ac is rated at 1.3Gbit/sec or 1300Mbits/sec (which is 300Mbits faster than 1Gbit Ethernet) (doesn't that beg the question, why doesn't the AP have a 10Gbit Ethernet port?)

 

In addition to the 802.11ac NIC built into my laptop, I also seprately tested (with built-in Wireless NIC disabled) using a USB 3.0 Adapter that is 802.11ac 1300Mbits, same results.

 

Slightly related note, any suggestions for a quality USB 3.0 802.11ac 1300 WiFi adapter compatible with Windows Server 2012 R2?  I just tried the ASUS AC1900 USB-AC68, but the driver installation fails to install the GUI to configure wireless networks inside Windows (though the driver and wireless network setup GUI works fine/as expected in Windows 10).

 

Your thoughts.

 

TIA!

 

---

You should type "show ap association" on the commandline of the IAP and see the capabilities of the client that is connected under the "phy" column.  Alot of performance also depends on multipath, so if the client has direct line of sight to the access point, it could perform worse.  

What is your iperf statement?

I updated my original post to include more information.

BTW, my test lab is in a relatively isolated building.

RF Channel Utilization: 1%

Noise Floor: 97 dBm

Error/Retries/Drop statistics: 3 fps

Neighbouring Clients: 1

Thanks for your thoughts!  I provided a lot of details about my benchmark environment so others can check their lab results too.  Pls add your performance stats to this thread.

Note that 1300Mbps is the on-air rate, wired data throughput theoretical max is far less. You would see max performance rates in the high 700s to low 800s Mbps.  

You are seeing approx 340mbps out of your USB wireless NIC which is about par for the course in most USB NIC testing I've done and seen. 

You can pretty much ignore any displayed rate in windows as it's a measure of connected rate and rarely ever reflects actual link rates. 

If you can test with a 3x3 11ac wireless laptop that has an on-board/PCI NIC, that should net you better results. But USB NICs are usually much slower than their PCI/laptop counterparts.

Jerrod, thanks for your experienced feedback.

While I don't have a laptop with a PCI 802.11ac 3x3 NIC, I was able to get an ASUS PCE-AC68 802.11ac 3x3 PCI-e card into a dual Xeon E5 server running Windows Server 2012 R2.

Even with all that power, transfer rates were slightly slower than the ASUS USB NIC which was benchmarked using the same server hardware.

FTP 2.3GB compressed file transfer = 456Mbits/sec

iperf -c server -t 10 -P 10 -w 64k = 331Mbits/sec

So, no speed increase using a PCI-e adapter card over USB 3.0.

Note: Both the Asus USB and PCIe adapters tend to lock onto 2.4GHz first, so I purposely configured my iAP-215 for 5GHz only.  However, then Windows would not see the SSID at all.  I found it necessary to restart both my Windows computer and the iAP-215, then they could see each other.

Jerrod, of the 1300Mbits of "air rate" (will you explain that please), in the situation when one computer is downloading at maximum real world speed as indicated in my testing (let's say 456Mbits/sec), is there simultaneous unused available bandwidth for another computer to simultaenous transit data without interfering with the first download?  IOW, without sharing the bandwidth?  I'm curious about the allocation of that 1300Mbits.  Can you break down the working parts and sections into understandable terms, please?  For example, 10% is used for signaling, 5% for WMM, etc.

Oh, regarding the iAP-300 series, which appears to add a 4th stream, what real world transfer rates does it have?

I saw approximately an 80Mbits/sec difference between my 802.11ac 2 stream and 802.11ac 3 stream.  Since 3 stream is about 456Mbits/sec, is each stream worth about 152Mbits/sec?  So, 4 stream will be 608Mbits/sec?

Thanks!

I would start here

http://www.arubanetworks.com/pdf/technology/whitepapers/WP_80211acInDepth.pdf

Then go here for the 11ac Wave 2 deck to answer your question about the 4th stream and it's impact.

http://community.arubanetworks.com/t5/Airheads-Conference-Vegas-2015/Breakout-802-11ac-Wave-2-technology-deep-dive-and-deployment/td-p/218559

Those two docs should cover pretty much any question regarding the 'where does that value come from'. Remember the client devices are more the limiting factor here (ala there are no four-stream clients out today that we have found). And MU-MIMO adds a whole new dimension to the equation.

The 1300Mbps is the max on-air PHY rate for VHT80 (in the example you've given, that value is primarily impacted by the AP radio characteristics such as number of streams, max channel width, etc which is all covered in the two docs above). It's wireless so wifi being half-duplex, the real-world throughput is some value less than that (ballpark can be that it's half though it's not that cut and dried as it depends on load and test config). As far as how it breaks up, it's ALL a shared medium. If you are running your AP on Ch36E VHT80 (So channel 36-48), then that spectrum (channel+channel width) is a fixed value shared with any and all clients on that AP at any given time. How that breaks up, which clients get what percentage of the bandwidth, etc is all completely determined by what the clients are trying to do. A frame is a frame so you can have situations where a voice client passing all very small frames can tie up the channel more than another client doing ONLY data passing the largest on-air frames as possible. Additionally, the protocol of the traffic matters as well (TCP versus UDP versus Multicast, etc). And finally, how you have configured your WiFi (did you trim any of the lower data rates, any special wifi capabilities added like 11r/11d/11?), are all the clients the same (1 stream versus 2 stream versus 3 stream verseus 11n and 11ac or even some 11g clients, etc) all come in to play. So it's not somethign you can generally break down into an easy calculateable formula. 

You also need to be clear when you quantify your performance. 57MBps is equal (roughly) to 456Mbps (which is not that terrible). Your first post you were using MB and in the latest you are using Mb, and there's a huge difference (standard is to use Mbps). 

Also if you are testing with USB WiFi Nics, you should expect that the throughput will never get to the level of the maximum as the USB bus adds overhead to the transaction that prevents the same performance levels as what you find with SOC or on-chip/PCI connected wifi cards. 

Thanks for all that informative detal. I'll check it out.  I updated my posts to reflect only MBits/sec.

Here's the Asus 4 Stream PCIe Adapter.  Yes, it supports MU-MIMO technology.

http://www.asus.com/us/Networking/PCE-AC88/

I can't afford to spend more time benchmarking, but if you do a 3 stream vs 4 stream test with all other considerations static, I would love to hear the results!  Although, if someone sends me a iAP-320 (hint, hint), I'll repeat my testing!

Curious, what does it take to get a vaunted 160MHz connection?  Which Aruba equipment and what type of client equipment (4 stream)?  Is 160MHz even possible with a iAP-215?

VHT  : Very High throughput; 80: 80MHz; 160: 160MHz; 80p80: 80MHz + 80MH

This is not an adapter that we would usually do any benchmark testing on, as it's not a typical wifi client device (laptop, phone, tablet, etc), but might be interesting to get one of to do some general testing with. For all other testing we do though, keep an eye out for our blogs as we usually post competitive testing there, where we look at single-client, 30, 60, and 100 client testing results. That's a good source for most of our benchmarking results. Thanks! 

So a quick update. I pinged our competitive manager to mention this 4x4 client and they had already tested it against out AP-335 (essentially our top of the line AP against the Asus 5300, their top of the line AP). Results are:

Observed that the PCE-AC88 was not any better performing as a 4x4 11ac Wave 2 client compared to the MBP 3x3 11ac client, both on our AP-335 and the Asus 5300. So on the face of it as a client there's likely not much of any speed benefit for 4x4 versus 3x3. 

They also did single and multi-client testing. All client MBP (3x3 11ac) Results:

AP-335:

Single Client = 840 Mbps

50 Clients = 510 Mbps aggregate

Asus 5300:

Single Client = 805 Mbps

50 Clients = 80 Mbps aggregate

Speaks to the difference between enterprise APs and SOHO/Home APs, even at the high end. 

Single 

Single Client

Great.  Thanks for your extra effort! :-)

Sorry, I don't recognize "MBP 3x3 11ac client".  Is that an Aruba product?  Part number?

MacBook Pro (Apple) which is a 3x3 11ac Wave 1 client.