How often are you asked the following question: Have you tried rebooting? Worse still, how often do you ask this question of someone who is asking you for help? Software rejuvenation--in euphemistic language--is the process of rebooting a system when things go wrong.

The authors discuss the situation where the cause requiring the reboot is a malicious piece of code, with worm-like properties replicating it throughout the network; the worm replication rate is constant. Another major simplification is that infection only causes performance degradation and not incorrect functioning. However, the authors do allow for periodic performance degradation due to garbage collection, to ensure that infection detection is not absolutely trivial.

Given the above scenario, the authors evaluate a few algorithms for triggering rejuvenation. Their ultimate algorithm uses variable depth buckets for monitoring transaction processing times, two trigger levels (depending on if the node is in a mission critical state), and warnings to adjacent nodes whenever rejuvenation takes place. The authors’ simulation results show the incremental improvements in mission success that each of the above three aspects provides under varying loads. With the simulation results, the authors develop an analytical model that predicts the probability of mission success given a probability for node infection.

In many respects, the simulations are far too simplistic to be representative of the real world, to which anyone who has tried a reboot to get rid of a virus can testify. However, overall, this is an interesting and readable paper.