Scalable Load Balancing in Networked Systems: A Survey of Recent Advances

In this survey we provide an overview of recent advances on scalable load balancing schemes which favorable delay performance and yet require minimal implementation overhead. The basic scenario involves a single dispatcher where tasks arrive that must immediately be forwarded to one $N$ single-server queues. join-the-shortest-queue (JSQ) policy yields vanishing delays as grows large, in centralized queuing arrangement, but prohibitive communication burden. contrast, JSQ($d$) assign incoming task server with the shortest queue among $d$ servers selected uniformly at random little communication, lead constant delays. order examine fundamental trade-off between overhead, discuss body research JSQ($d(N)$) diversity parameter $d(N)$ depends investigate growth rate required match optimal JSQ fluid diffusion scales. Stochastic coupling techniques scaling limits play instrumental role establishing asymptotic optimality. We demonstrate how methodology carries over infinite-server settings, finite buffers, multiple dispatchers, arranged graph topologies, token-based such join-the-idle-queue (JIQ), thus providing broad main trends field.