4/19/2013

The State of the Art: Increasing Network Resiliency by Optimally Assigning Diverse Variants to Routing Nodes (a reflection)

Cristina Nita-RotaruAssociate Professor in the Department of Computer Science at Purdue University, presented a talk at Depaul University on research efforts currently underway at Purdue's Dependable and Secure Distributed Systems Laboratory

The following is a reflection on some of what I took away from the talk. A high-level overview of the research, presented visually, can be seen by clicking here


Summary


Within cloud computing, the Network as a service (NaaS) framework is uniquely suited for deployment within a security model based on intrusion tolerance, yet the majority of NaaS deployments currently rely on intrusion detection mechanisms instead. Defining a protocol to support such a model is a major area of research at the laboratory, and thus is where Nita-Rotaru focused her talk.
  • Research Questions: What is the optimal way to maintain the performance and accessibility of distributed networks whose topography may expose critical data to hostile environments?
  • Hypotheses: Strategic use of diverse variants (e.g., various operating systems) within distributed networks will provide increased performance and greater resiliency than that provided by any model currently in use. 
  • Methods: Simulations were used to test various configurations of diverse variants across a wide spectrum of potential environments.
  • Conclusions: Results showed that by increasing diverse variants, network performance and resiliency could be increased if said variants were optimally configured. They determined that if at most three (3) variants are arranged at nodes across the network in such a way that any given node had equal exposure to more than one (1) variant at any given time, resiliency increased by as much as (approximately) 15%. What was surprising to the team was the discovery that suboptimal configurations could actually degrade network performance and resiliency. Although seemingly counterintuitive, this phenomenon can be explained as follows; for any random assignment of n variants, the likelihood of reaching an optimal state is far less then the likelihood of reaching a sub-optimal state. 

Connections


There is a degree of inherent robustness afforded by a corresponding degree of natural diversity, which animates all of the dynamic living systems on our planet. What is so fascinating about this research is that the same appears to be true for the non-natural dynamic systems that we humans have artificially created. Perhaps the distinction between such networks is not as firm we perceive it to be.

Further, it could be argued that our human-crafted systems are in fact natural systems in their own right, insofar as they arise from our efforts in an iterative, unfolding process as dynamic and interdependent as any naturally occurring one. Is it not true for example, that the expansive underground networks built by ant colonies arise in a manner not dissimilar to our own? That is, as ant-crafted systems born from the natural action and movement of natural ant bodies as natural solutions to their problems. 

What delineating factor exists to separate the formative essence of these two acts of collectivist creation? As humans, our inclination is to suggest that it is our ability to be self-aware, to independently direct the arc of our consciousness, that is the delineating factor here. Perhaps, but science has yet to provide us a means to empirically prove the existence of such a delineation, and has even failed to provide a comprehensive understanding of the meaning of consciousness. 

I propose that similarities between our human-crafted networks and those present in nature far outweigh any differences. The information networks that humans have created thrive on diversity similar to the networks of ant colonies; flocks of birds; schools of fish; and other systems whose formation is driven by the actions of sentient beings and non-sentient processes alike. 

It is understood that the process of evolution is in part fed by environmental stresses applied over time, causing the system to segment and mutate in a process that sheds weaker segments while providing enhanced functionality to the already more robust segments, which in turn allows them to expand outward to increasingly diverse environments, environments which again stress the system in new ways, and so on. 

Indeed what I have described reads much like the ongoing story of the Internet and of the World Wide Web that it supports.