The growing autonomy of servers may significantly deteriorate the performance of traditional load-balancing strategies. Indeed, the authoritative decision belongs to the loadbalancer, but the autonomous servers may reject the requests on their own convenience. We propose in this paper an original loadbalancing strategy for transferring this authority from the loadbalancer to the autonomous servers. We describe the underlying architecture and evaluate our solution based on a first set of experimentations.

SUMMARYRPL is a routing protocol for low-power and lossy networks. A malicious node can manipulate header options used by RPL to create topological inconsistencies, thereby causing denial of service attacks, reducing channel availability, increased control message overhead, and higher energy consumption at the targeted node and its neighborhood. RPL overcomes these topological inconsistencies via a fixed threshold, upon reaching which all subsequent packets with erroneous header options are ignored. However, this threshold value is arbitrarily chosen and the performance can be improved by taking into account network characteristics. To address this we present a mitigation strategy that allows nodes to dynamically adapt against a topological inconsistency attack based on the current network conditions. Results from our experiments show that our approach outperforms the fixed threshold and mitigates these attacks without significant overhead.

No abstract