[OLSR-users] Question about OLSR

[Roar Bjørgum Rotvik]

ap wrote:
> Gecko64 wrote:
> 
>> Hello, i've got a question about OLSR.
>> I've test this protocole with 3compunters (compunter A B and C).
>>
>> A saw B and B saw C but C didn't saw A and when i've test in 11mbps, 
>> i've see a ping of 30ms between A and B and the same between B and C 
>> and i had a total ping of 60ms.

[...]

>> 60ms is not a lot yes but in the network than i'm building in my 
>> country, i'll have sometime 50 nodes between 2 compunters :-s
> 
> if you have a mesh network with a diameter of 50, you will most likely 
> have more than 2000 nodes running. i don't think olsr, or any other 
> existing MANET implementation, can handle that. your bandwidth will be 
> saturated by control messages alone.

Not only that, but you must remember that standard WLAN today only have 
one antenna, so it cannot send and receive at the same time.
So when B is receiving a ping from A, it must receive the full ping 
packet (maximum MTU size) before the packet is routed out on the WLAN 
interface again to be delivered to C. In practice this cut the bandwidth 
in half.

But if you have many nodes after another (A-B-C-D-E-F), and B and E 
doesn't hear another, then B may send to A or C at the same time as E 
send to D or F.

For 802.11b you get a practical troughput of about 4-5Mb for one link 
(A-B), and packets from A-C get half of that throughput, i.e. 2-2.5Mb. 
If these packets are forwarded further from C-D and both C and D are in 
hearing range of B, then the throughput from C-D are cut in half again, 
resulting in a throughput of approx. 1Mb from A-D (then add collitions 
on the air and so on). That's approx 125KB (Kbytes) bandwidth over 3 hop.
If another traffic from A-B or B-C that saturates the WLAN-link takes 
place at the same time, traffic from A-D would suffer a lot and perhaps 
stall completely.

But this is dependand on the coverage range of each node and other WLAN 
problems like hidden-node, mutipath reflections and so on.

So I would argue that a 50 hop standard WLAN network seems not very 
likely in practice. I remember using the ad hoc routing solution Lunar, 
that had a hardcoded limit of 3 hop, because more hops was difficult in 
practice.

I know that some on this list have real OLSRD WLAN network deployed, 
what is the highest hop count used in practice today?

>> The question than i'm asking was if it was possible to have a ping of 
>> 30ms between A and C? If it's possible to optimise this lag... or if 
>> it existe an other protocol which one i'll not have this problem?
> 
> the delay is most likely NOT caused by olsr. olsr just builds the 
> reouting tables. afaik, the actual forwarding of packets is handeled by 
> your operating system. switching the routing protocol will not improve 
> the performance.
> to improve the ping times i would suggest the following:
> 
> 1) try 54 mbps hardware.
> 2) use a processor with at least 200 mhz.
> 3) use an operating system that has a good routing performance.
> avoid using windows 95/98/ME. use linux, or, if you are bound to 
> windows, use XP or 2000.

As I have mentioned above, the bandwidth from A-> C is cut in half 
compared to one hop link (A-B or B-C) because of how WLAN works today 
(one antenna), so while 1) would yield lower ping time, the ping time 
from A-C cannot be as low as ping time between A-B or B-C.

-- 
Roar Bjørgum Rotvik