Dynamic overload balancing in server farms

Abstract

Abstract-We consider the problem of optimal load balancing in a server farm under overload conditions. A convex penalty minimization problem is studied to optimize queue overflow rates at the servers. We introduce a new class of α-fair penalty functions, and show that the cases of α = 0, 1, ∞ correspond to minimum sum penalty, penalty proportional fairness, and min-max fairness, respectively. These functions are useful to maximize the time to first buffer overflow and minimize the recovery time from temporary overload. In addition, we show that any policy that solves an overload minimization problem with strictly increasing penalty functions must be throughput optimal. A dynamic control policy is developed to solve the overload minimization problem in a stochastic setting. This policy generalizes the well-known join-the-shortest-queue (JSQ) policy and uses intelligent job tagging to optimize queue overflow rates without the knowledge of traffic arrival rates.

Limiting Properties of Overloaded Multiuser Wireless Systems With Throughput-Optimal Scheduling

“…Such a greedy allocation can also be interpreted as that the scheduling algorithm does not consider the fairness at all. When α → ∞ the problem approaches minη ∈C λ |λ −η| ∞ , i.e., minimizing the L ∞ norm of vector λ −η, which means the queued traffic satisfies a minmax fairness [34], the dual of the max-min fairness [32], [33]. From the above two asymptotic cases we can conclude that, the fairness of the blocked traffic can be guaranteed by choosing a proper α.…”

“…Furthermore, the author showed that, for a specific backlog growth direction, by designing a proper b i , the backlogged traffic is also minimized. In [34], the author discussed how to minimizing an α-penalty function of the queue overflow rate in a server farm. It turns out that by choosing ∀i, b i = 1, the problem (7) is a special construct of the problem discussed in [34].…”

“…In [34], the author discussed how to minimizing an α-penalty function of the queue overflow rate in a server farm. It turns out that by choosing ∀i, b i = 1, the problem (7) is a special construct of the problem discussed in [34]. However, as [24] and [34] both using identical α i , i.e., only discussed the fairness within each group, their works are unable to find the critical impact of α i on the fairness.…”

A scheduling framework for performing resource slicing with guarantees in 6G RIS-enabled smart radio environments

“…Particularly related to the present work is the negative drift concept and has been applied in multiple ields like mechanical engineering [33]. Related analysis regarding stochastic analysis tools (namely Lyapunov analysis in systems where negative drift is applied to change priorities) were used in several pieces of work [34], [35], [36]. In [37] and [38] aggregate packet rate guarantees to individual clients is combined with negative drift and measurement-based admission control techniques.…”

Service differentiation in virtual service - based network architectures

“…were used in numerous works like [56], [66], [67], [68] and [69]. Relevant work has also been presented in works like [70] and [70], where the Lyapunov optimization technique is performed.…”