hello guys, I'm facing a vexing problem...one of my colleagues is writing a network mgmt system thats queries our WAN devices, collects info on their 'neighbours' and is supposed to plot the topology of the network. so basically...the input to the plotting program would be a list of: nodeID,{set of its neighbours}
On Nov 15, 2002 at 02:02, kishor bhagwat wrote:
be a list of: nodeID,{set of its neighbours}
From this, it has to graphically plot the links.
I'll give an example of the issues...how do we know which node should be at the center? how about redundant links between the same two locations?
Look for ISI's networks simulator and network animator (ns and nam), I think the source code is available. At worst, you can interface with it using C++ and oTCL.
At 02:02 morn 11/15/02 -0800, Kishor wrote:
hello guys, I'm facing a vexing problem...one of my colleagues is writing a network mgmt system thats queries our WAN devices, collects info on their 'neighbours' and is supposed to plot the topology of the network. so basically...the input to the plotting program would be a list of: nodeID,{set of its neighbours}
From this, it has to graphically plot the links. I'll give an example of the issues...how do we know which node should be at the center? how about redundant links between the same two locations?
But does a network necessarily have a center? In the sense that I did not understand what you mean by a center. Anyway the plot will be relative. So suppose one WAN device is the center of some type of star topology then it will have multiple links with different devices with the help of which you can determine its 'centerness'. Some redundant links will overlap and others could be mostly weeded out by a little external knowledge I think. But I am not sure I understand you fully.
Does anyone know of any mathematical algorithms for solving such a problem? well..would be glad if anyone can help!
Graph theory deals with this. But I have not understood your problem enough to see any need for any such algorithm... Looks like an interesting problem though, and we can continue it off list if it so interests you.
cheers, quasi
The geographical location of the item on the network map generally starts from the node where the mapping software is running. The other objects on the network should then be mapped based on whether there is a direct link from the starting node to it or goes thorough another node.
eg : from my pc, there is no direct connection to node 4, it goes through a hub. So the my pc and node 4 will be centered round the hub. Similarly, if the connection goes through the same hub and then through a switch to go to node 105, the switch then is the center for all hubs, nodes connected through it. and so on
Regards Saswata
----- Original Message ----- From: "q u a s i" quasar@vsnl.net To: linuxers@mm.ilug-bom.org.in Sent: Saturday, November 16, 2002 12:18 AM Subject: Re: [ILUG-BOM] OT - Topology plotting
At 02:02 morn 11/15/02 -0800, Kishor wrote:
hello guys, I'm facing a vexing problem...one of my colleagues is writing a network mgmt system thats queries our WAN devices, collects info on their 'neighbours' and is supposed to plot the topology of the network. so basically...the input to the plotting program would be a list of: nodeID,{set of its neighbours}
From this, it has to graphically plot the links. I'll give an example of the issues...how do we know which node should be at the center? how about redundant links between the same two locations?
But does a network necessarily have a center? In the sense that I did not understand what you mean by a center. Anyway the plot will be relative. So suppose one WAN device is the center of some type of star topology then it will have multiple links with different devices with the help of which you can determine its 'centerness'. Some redundant links will overlap and others could be mostly weeded out by a little external knowledge I think. But I am not sure I understand you fully.
Does anyone know of any mathematical algorithms for solving such a problem? well..would be glad if anyone can help!
Graph theory deals with this. But I have not understood your problem enough to see any need for any such algorithm... Looks like an interesting problem though, and we can continue it off list if it so interests you.
cheers, quasi
Saswata Banerjee & Associates writes:
The geographical location of the item on the network map generally starts from the node where the mapping software is running. The other objects on the network should then be mapped based on whether there is a direct link from the starting node to it or goes thorough another node.
eg : from my pc, there is no direct connection to node 4, it goes through a hub. So the my pc and node 4 will be centered round the hub. Similarly, if the connection goes through the same hub and then through a switch to go to node 105, the switch then is the center for all hubs, nodes connected through it. and so on
Regards Saswata
This is a typical problem in graph theory. Maybe be a linked list with multiple links may solve the problem. but the biggest problem is that of cycles in the graph which have to be handled. Also another problem is that you have to know beforehand what is the maximum number of nodes that a particular node is connected to, unless there is some way to get around this problem.
************** Vinayak Hegde APGDST Student NCST-Juhu **************
--- Saswata Banerjee & Associates scrap@saswatabanerjee.com wrote:
The geographical location of the item on the network map generally starts from the node where the mapping software is running. The other objects on the network should then be mapped based on whether there is a direct link from the starting node to it or goes thorough another node.
not really...if i'm a vendor managing someone's network..i'll have the software running in my data center..and the network spread across the globe. Secondly..the software could be installed anywhere in the network itself. The concept of center is needed to propoerly layout the nodes on the screen..Think of the nodes lying on concentric rings.If i've already plotted a node, and it appears again as another node's neighbour, i need to know at which ring it is...
regds, kishor
__________________________________________________ Do you Yahoo!? Yahoo! Web Hosting - Let the expert host your site http://webhosting.yahoo.com
At 10:09 even 11/16/02 -0800, Kishor wrote:
not really...if i'm a vendor managing someone's network..i'll have the software running in my data center..and the network spread across the globe. Secondly..the software could be installed anywhere in the network itself. The concept of center is needed to propoerly layout the nodes on the screen..
Wouldnt a bit of human-int help here? I mean could you not define a node, say the company headquarters or main router etc.etc.. as your 'center' with all the others relative to it.
Think of the nodes lying on concentric rings.If i've already plotted a node, and it appears again as another node's neighbour, i need to know at which ring it is...
If A is neighbour to B then is B too a neighbour to A ? I mean if you are just mapping the entire network devices then it should be... but in the case you are also considering the traffic routes then maybe not, right? Concentric rings in what sense? Levels? i.e. first order neighbour to center, second order neighbour to center.. etc.etc? Then maybe you could add another field directly to the node saying at what level it lies. This may make it easier to locate it on one of the concentric rings... In short you define a starting point for the software which then explores all the network determining the level of each node so as to plot the relative positions.
quasi
Er, why not check the ns source code, or even think about writing a TCL script and using ns/nam directly?
On 15/11/02 02:02 -0800, kishor bhagwat wrote:
hello guys, I'm facing a vexing problem...one of my colleagues is writing a network mgmt system thats queries our WAN devices, collects info on their 'neighbours' and is supposed to plot the topology of the network. so basically...the input to the plotting program would be a list of: nodeID,{set of its neighbours}
From this, it has to graphically plot the links.
I'll give an example of the issues...how do we know which node should be at the center? how about redundant links between the same two locations?
Does anyone know of any mathematical algorithms for solving such a problem? well..would be glad if anyone can help!
cheops? Older versions had source available, before it went closed source. I remember picking it up from the.wiretapped.net
Devdas Bhagat
-
Devdas Bhagat -
kishor bhagwat -
q u a s i -
Saswata Banerjee & Associates -
Satya -
vinayak_hegde＠softhome.net