OSLR overhead question.

[David L. Vrablic]

Tog and the gang that are using OSLR.
Can you give me any idea of the overhead CPU usage in Typical % for a busy network.?
This network has about 30 radios and I will only be able mesh about 10 of them to start
.
The time has come to set up some multi paths for this network.
Keeping track of the routing is killing us when we add anything more.
I was just wondering where the point was -where I would get into loading problems.
All units in the system are 533 WAR 4 or AU 400 units.

;) Any hints so we don't fall into a hole along the way?

(Austin and Lenny you had better read this and heed their advice.)
(There are some mighty smart people on this forum!)

[Beebe]

My board running at my house is about 106 feet overhead.

Thanks,
Roger

[Beebe]

Sorry - I'll get my coat.

[David L. Vrablic]

You be funny man! Beebe.
I need some humor right about now.
Thanks!

[tog]

To answer your question, the extra load is negligible. I didn't even notice any extra CPU usage with OLSR and I have over 100 routes listed at the moment, I just looked.

To give you a more concrete idea, my core router is currently running on an Athlon 1GHz and olsrd 0.5.0 has been up for 50 days. It has used 325 minutes of CPU time in that time, or about .0045 minute of CPU time per minute, or about 0.45% of the CPU on average. FreeBSD actually only shows olsrd taking 0.00% CPU with an occasional 0.05%.

So hopefully that gives you an idea of what I mean by negligible. Even if it's double that on a 533MHz RISC processor it's still pretty negligible.

[David L. Vrablic]

That is great news.
I have been studying various meshed protocalls and all of the early ones were plagued with horrendious CPU overhead demands.
I finally settled on trying the OSLR and there wasn't a peep anywhere about the CPU impact.
Tog, Yours is the first report from any quarter.
I have read about some of the major cities tearing their systems down and starting over or just abandoning the whole concept of city wide WIFI.
I have a hunch some of it was driven by hard cold fact that someone has to pay in the end and ongoing expenses were not accounted for.
The rest found out they couldn't do it with 1 radio or even 2 at each node.
All my nodes have 4 radios except for out on the very ends.
It is now time to connect the dots and make a fully meshed system.
I am just trying to plan ahead.
I am very pleased with the scaleability of this gear.
About 30 units on this camera system and it isn't even breathing hard.
Gotta love it when something works.
I wonder how many I can add or how far I can turn it up before it breaks down?
I need to beef up some of the point to point paths and add the mesh then we will find out. Right now 8FPS at 640 x 480 uses about 1.3 Meg per camera. As soon as the SNMP is fixed I can tell more for sure.
Thanks guys This is fun!

[Beebe]

Here's a question. Does anyone have a reason why OSLR is in any way superior to OSPF? Shouldn't anyone who has migrated past a DSL connection be using OSPF and BGP if they're multihomed?

I'm honestly genuinly interested in finding a true answer to this, Isn't this the industry standard? It does seem to be the industry standard for BIG ISPs, but doesn't seem to be the industry standard for wireless ISPs and I was curious why?

Anyways, to go back to the thread topic (kinda), why was OSLR even included in staros, and why is it any better than OSPF?

Thanks,
Roger

[lonnie]

Roger, I am tempted to answer this, but then everybody would know why we chose OLSR. I am sorry, but we have to let the other guys do their own homework.

We did a lot of testing and evaluation and OLSR satisfied us from a technical and operational level. OSPF might be the Industry Standard but it has always been quirky.

[Beebe]

Does OSLR work with multiple default gateways? Does it just route packets to the one with the lowest hop count?

I'll probably switch to it since I've seen the "quirkyness" you mention with OSPF over wireless. I have to upgrade a few more boards on the network first though, get rid of the last few prism/v2 APs.

Thanks,
Roger

[lonnie]

OLSR chooses the best path, as determined by hop count and tweakable metrics you can set to give a preference of one path over several otherwise identical paths.

[valenti]

There was some discussion of olsr on a wireless list a few weeks ago. (either
the WISPA General Wireless List: wireless@wispa.org or Archives: http://isp-lists.isp-planet.com/isp-wireless/archives/ )

The comment that I remember is that olsr was designed to recognize erratic link quality and compensate. The entire post was very good, I even printed it off for reference - of course it isn't here. It gave me a warm, fuzzy feeling about using olsr (in the future).
----
Ah, here it is: message from Clint Ricker, Subject Re: [WISPA] Locustworld meshes? dated Sept 15, 2007. I would copy it here, but maybe that isn't proper and it should be at Archives: http://lists.wispa.org/pipermail/wireless/