06-20-2012 05:32 PM
We use quite a few different RAPs (AP61, AP93H, RAP5, AP125), and we've noticed that WAN throughput performance generally hovers around 50% of available bandwidth.
For example, connecting directly to a wired connection gives a FTP throughput of say 14Mbps.
Where connecting to a RAP doing split tunneling (DHCP over tunnel, everything else routed locally), our bandwidth drops to 7Mbps.
Switch the connection to a 100Mbps wired connection, and again, direct connection yields 80Mbps while the RAP gives 20-30Mbps.
This is consistent across different RAPs, and we have ruled out broadcast traffic, ARM, or anything that looks like it would get in the way. I'm currently testing on a lab setup to rule out any other factors that could affect this, and I am the only client connected to the test RAP.
I see quite a few threads about similar issues but I don't see many solutions offered other than calling the TAC.
Are there specific configuration areas I should be focusing on to help diagnose and resolve this issue?
My lab is currently running 126.96.36.199.
06-20-2012 07:05 PM
You should be getting much more on the RAP-5 than on the other platforms you are mentioning depending on the variables you are using in the tests. Then again I am not clear exactly how you are testing.
You mention "Wired connection", which I take to mean a 100 Mbps, full duplex wired NIC connection. Correct? That seems straightforward.
So... here is the unclear part: What are you using to benchmark against that? Wireless ? Wired? Combination? I ask as you mention some APs that have wired ports (so you could be doing wired vs. wired), then again you mention the Ap-61 (which outside of being a real old AP that I wouldn't expect to see in use very often if at all any more(given it doesn't do 802.11n) doesn't ahve wired client connectivity persay....
A few more details would be great to have so we can discuss more in detail.
Curious, in how many sites do you have 30 Mbps+ WAN connectivity for the RAPs to connect to ? I know bandwidths are getting cheaper and faster... but the norm is < 10 Mbps in the vast majority of cases I see in practice... what are you guys seeing?
06-21-2012 11:48 AM
That's what I thought as well... I should be getting much more performance than what i'm seeing.
And the drop is very consistent.
We use different setups for testing and deployment: LTE connections provide real world throughput of 30Mbps/20Mbps with 30ms latency.
Our office has 100Mbps symmetrical fibre with 5ms latency.
On both connections, we see the drop. I'm comparing this to connecting my laptop directly to the wired connection, as well as connecting a dumb (linksys, etc..) wireless AP to the same connection.
On top of this, i'm playing with the forward mode configuration, and the throughput follows the pattern I would expect: tunneled (slowest), split-tunneled (mid), bridge (fastest), but they're all still losing 50%+ of the total known throughput.
I included the AP61 in my results, simply because I wanted to see if it exhibited the same behaviour. In reality, we deploy AP93/105 etc... exclusively.
11-28-2012 05:21 PM
Where you able to fix this?
I was testing something similar... and i got a tread of it...
Product Manager - Aruba Networks