אירועים והרצאות בפקולטה למדעי המחשב ע"ש הנרי ומרילין טאוב

Reliability and time-efficiency are two key elements to consider in network design. Commonly, each is measured per service - availability probability of a specific service, the latency of a specific service, and overall - system average reliability and system average latency, considering the demand for every service. Intuitively, minimizing latency requires minimizing the number of network elements a service makes use of. In a non-redundant environment, this would also guarantee the maximal reliability of a service, as reliability degrades when the number of elements required to implement the service increases. However, reliability is often guaranteed by allocating redundant, backup resources. We explain that such redundancy or the joint support for multiple services can impose a trade-off between reliability and time-efficiency criteria. We show in what cases an optimal solution to the latency problem is an optimal solution to the reliability problem and vice-versa. For the cases in which the trade-off exists we examine if we can bound the distance to an optimal solution of a given criterion (latency/reliability) using the optimal solution of the other criterion. Our analytical model considers networks following a virtualized paradigm - network functions are implemented on commodity servers and a network service is implemented by chaining functions.

*M.Sc student under supervision of assistant Prof. Ori Rottenstreich