Heaing/Repair/Resilience in Networks( Computer networks, Network models): Difference between revisions
From Santa Fe Institute Events Wiki
Alexhealing (talk | contribs) No edit summary |
No edit summary |
||
| Line 9: | Line 9: | ||
[[User:Alexhealing|Alexhealing]] 18:39, 5 June 2007 (MDT) | [[User:Alexhealing|Alexhealing]] 18:39, 5 June 2007 (MDT) | ||
== Overview == | |||
My idea of proposing this discussion is to look for projects/ideas (like those indicated by Alex above) in the broad area. The problem is how can networks survive/repair/maintain functionality under certain attacks/losses. This is a very general idea which can be adapted to many fields and models. Here's an overview of my recent research (I will hopefully put a conference abstract on my homepage soon): Many networks use overlay networks, redundancy etc. for resilience. Our idea is that in certain kinds of networks called reconfigurable networks (peer-to-peer networks, ad-hoc networks, social networks etc.) we need not put redundancy beforehand (or have limited redundancy). Upon an attack (deletion of a node), the neighbours of the deleted node reconnect themselves to maintain connectivity in the network (this is 'self-healing'). The quality of healing can be described by stating if the system can maintain some invariants/ keeping resource costs low, like keeping the degree increase of any node within a certain limit, or keeping the diameter small. We have been successful in making a distributed/localized selfhealing algorithm which can guarantee degree increase of no more than O(log n). But there are many directions to expand the idea into -> deletion of whole clusters, maintaining both degree and diameter, selfhealing in networks which compute some functions (the brain, Electric circuits) etc. | |||
[[Amitabh_Trehan]] | |||
Latest revision as of 21:01, 6 June 2007
I'd certainly be interested in having a discussion related to my work involving the engineering of complex adaptive agent overlays to autonomically manage the information layer of networks for decentralised computation.
I'm also particularly interested in learning more about biological networks (e.g. GRNs) and discussing possible analogies with computer networks which may harness any of self-* capabilities such as
- self-healing
- self-configuration
- self-optimisation
- self-organisation
- self-protection
Alexhealing 18:39, 5 June 2007 (MDT)
Overview
My idea of proposing this discussion is to look for projects/ideas (like those indicated by Alex above) in the broad area. The problem is how can networks survive/repair/maintain functionality under certain attacks/losses. This is a very general idea which can be adapted to many fields and models. Here's an overview of my recent research (I will hopefully put a conference abstract on my homepage soon): Many networks use overlay networks, redundancy etc. for resilience. Our idea is that in certain kinds of networks called reconfigurable networks (peer-to-peer networks, ad-hoc networks, social networks etc.) we need not put redundancy beforehand (or have limited redundancy). Upon an attack (deletion of a node), the neighbours of the deleted node reconnect themselves to maintain connectivity in the network (this is 'self-healing'). The quality of healing can be described by stating if the system can maintain some invariants/ keeping resource costs low, like keeping the degree increase of any node within a certain limit, or keeping the diameter small. We have been successful in making a distributed/localized selfhealing algorithm which can guarantee degree increase of no more than O(log n). But there are many directions to expand the idea into -> deletion of whole clusters, maintaining both degree and diameter, selfhealing in networks which compute some functions (the brain, Electric circuits) etc.