| RB1100AH Slowdown I bought this router thinking it would have enough horsepower to work as an edge router. That was quite some time ago.
I didn't find it to be stable enough to run Layer 7 protocols which is what I primarily use for traffic shaping.
As it sits now I have it doing NAT, about 7 simple queues that aren't even used for limiting. They have 20meg/20meg limits - basically in place so I can log how much data a few users are using. A few static NAT's for customers with static addresses.
Thats it. No layer 7, no QoS, no Mangle, no nothing. Very basic.
I have the CPU running at 1333mhz and it has 2gigs of ram.
If I connect directly to my upstream provider's Cisco switch I get 75megs/75megs. Sometimes it will go above 80megs. If I run it through the 1100AH with all interfaces disabled expect my laptop and the WAN port I get similar results. As soon as I enable ports going to my customers my throughput drops to about 6megs with it under 10megs of load.
I do have separate DHCP running on individual interfaces if that matters.
I don't see CPU spikes of any concern even when I'm running speed tests.
Any ideas? |
|
| Wow that's weird. There must be something wrong with the queues? I run a few of these and don't have anything like this. Only difference is I don't do any L7.
As far as I'm concerned, L7 sucks on any routerboard. It works well on a PC platform or probably a cloudcore, but RBs just don't have the kick needed.
That said though, your results are wicked low. Something just ain't right.
What about a complete re-flash and manually enter everything while watching the bandwidth? You can see the instant you trigger the speed drop and what is doing it.
Also, as usual, what firmware, and what routerboard firmware? |
|
| reply to wirelessdog
I kept trying to insert a RB450g and then a new 2011 something and each time they locked up. Even with all the rules disabled, the 2011 locked up.
Does yours happen to be in bridge mode?
I think these little guys just have a hard time seeing all that traffic at once and trying to figure it out (connection tracking).
I built a small dual core Atom that seems to be kicking arse at the moment with about 50Meg, layer 7, and other various rules working on it in bridge mode now.
I believe I tried the other two devices with 5.2 something versions. The x86 now in use, I believe has the latest stable for it installed.
Have you tried other firmwares?
--
»www.wirelessdatanet.net |
|
| reply to wirelessdog
I have the newest stable firmware running both ROS and RB firmware.
Are their settings that can be modified for connection tracking that might help? |
|
| reply to wirelessdog
I am bridging two interfaces but not the Interface I was testing on and not the WAN interface.
Routing. |
|
|
|
 TomS_Git-r-donePremium,MVM
join:2002-07-19
London, UK
kudos:4
1 edit | reply to wirelessdog
What are the speed and duplex settings for your MT and the Cisco switch?
Sounds like a classic case of duplex mismatch, i.e. one end being full duplex (possibly hard coded) and the other half duplex as a result of auto-negotiation failure.
Make sure that both ends are either auto-neg (preferable), or hardcode them in the same way.
edited |
|
| reply to wirelessdog
Confirmed speed/duplex auto negotiation all match. 1gig/full and 100meg/full respectively. |
|
TomS_Git-r-donePremium,MVM
join:2002-07-19
London, UK
kudos:4 | Config wise, or what they actually negotiate at? Because you cant have one side at 1G and the other at 100M... ;-P
Might need the provider to check the Cisco side, but you could also press the button on the front of the switch to cycle through the different LED modes to confirm on-site. |
|
| reply to wirelessdog
I have confirmed that one the remote side the upstream provider's switch is negotiated at 100/full with the Ubiquiti Rocket. On the near side I have confirmed that the upstream provider's switch is negotiated with the Ubiquiti Rocket at 100/full. My connection between the upstream provider's switch and my routerboard is 1gig full. These are actual negotiated rates confirmed on my equipment and the upstream providers switch. |
|
 warwick
join:2009-06-05
Hollywood, FL | reply to wirelessdog
Hmm, have you tried disabling the simple queues ( all of them )while troubleshooting?
ran into a similar situation about three weeks back where I changed or added a simple queue and the routerboard (450g - Core) essentially locked up. I.e. Ping Times of 400ms + and throughput no greater than 800kbps.
Disabled the offending queue and voila... back to normal. |
|
