On Sat, 27 Sep 2014, Nishant Sharma wrote:
Hi,
We have a batman-adv mesh network with 10 nodes (4 gateways) running on
Attitude Adjustment in a large warehouse (100,000 sq. ft.).
The problem occurs during peak hours when around 60 clients (laptops +
desktops) are connected, network slows down or grinds to a halt and
latency hits the roof. Ping starts dropping within LAN.
The problem occurs during the peak hours when all the stations are
connected and even if we plug two computers directly to the core switch
we get high latency and ping drops.
This actually sounds very typical for the way that a wireless network collapses
under load. This behavior is why the OLPC dream of using mesh networks for
everything ended up failing.
If you can, deploy more nodes with lower power per node to reduce the
interference. I'm not familiar with batman, but more nodes also gives you more
channels to play with (assuming you can use multiple channels)
I notice that you use 40MHz channels for 5GHz, unless you have wired connections
between the nodes, this seems like it's probably a waste as you only have 20MHz
channels on 2.4GHz. In theory it takes less time to transmit the data on the
40MHz channels, but in practice I'm not sure it really gains you a lot (it
definantly doesn't gain you as much as if you had two 20MHz channels operating
independently in that area)
I cover a lot of this in the talk I gave at LISA in 2012
https://www.usenix.org/conference/lisa12/building-wireless-network-high-density-users.
David Lang
Main applications used on the network are Windows File Share, RDP and
browsing.
Hardware used is ALIX2 with Atheros cards (Wistron DNMA92 - AR922x
chipset) using ath9k module.
Following are the relevant configurations:
batman-adv:
config 'mesh' 'bat0'
option 'interfaces' 'adhoc0'
option 'aggregated_ogms' '1'
option 'ap_isolation' '0'
option 'fragmentation' '0'
option 'log_level' '7'
option 'orig_interval' '5000'
option 'bridge_loop_avoidance' '1'
option 'distributed_arp_table' '1'
option 'network_coding' '1'
option 'gw_mode' 'server'
option 'gw_bandwidth' '100mbit/100mbit'
option 'bonding' '1'
network:
_snip_
config interface 'wan'
list ifname 'eth1'
list ifname 'bat0'
list ifname 'wlan0'
option proto 'dhcp'
option type 'bridge'
option hostname 'hopmesh-HB000DB9303058'
config interface 'mesh'
option mesh 'bat0'
option proto 'batadv'
option mtu '1546'
wireless:
config wifi-device radio0
option type mac80211
option channel 6
option hwmode 11ng
option path 'pci0000:00/0000:00:0c.0'
option htmode HT20
list ht_capab SHORT-GI-40
list ht_capab TX-STBC
list ht_capab RX-STBC1
list ht_capab DSSS_CCK-40
# REMOVE THIS LINE TO ENABLE WIFI:
option disabled 0
option country 'US'
option txpower '27'
config wifi-device radio1
option type mac80211
option channel 165
option hwmode 11ng
option path 'pci0000:00/0000:00:0e.0'
option htmode HT40-
list ht_capab SHORT-GI-40
list ht_capab TX-STBC
list ht_capab RX-STBC1
list ht_capab DSSS_CCK-40
# REMOVE THIS LINE TO ENABLE WIFI:
option disabled 0
option country 'US'
option txpower '27'
config wifi-iface 'wlan0'
option device radio0
option ifname wlan0
option network wan
option mode ap
option ssid blahssid
option encryption 'psk2+tkip+ccmp'
option key 'blahblah'
config wifi-iface 'wmesh'
option device 'radio1'
option ifname 'adhoc0'
option network 'mesh'
option mode 'adhoc'
option ssid 'blahssid'
option bssid 'AE:XX:B2:90:XX:XX'
option key '1'
option key1 's:blahkey'
We have bridged "wan" with "bat0" so that stations could get IP from our
main DHCP server for the network.
Could it be the ARP or OGM messages flooding the network which chokes
the backhaul?
Any pointers would be much appreciated. I would be visiting the
warehouse again tomorrow.
Thanks & regards,
Nishant
_______________________________________________
openwrt-users mailing list
[email protected]
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-users
_______________________________________________
openwrt-users mailing list
[email protected]
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-users
