Online algorithms for geographical load balancing

Abstract

Abstract-It has recently been proposed that Internet energy costs, both monetary and environmental, can be reduced by exploiting temporal variations and shifting processing to data centers located in regions where energy currently has low cost. Lightly loaded data centers can then turn off surplus servers. This paper studies online algorithms for determining the number of servers to leave on in each data center, and then uses these algorithms to study the environmental potential of geographical load balancing (GLB). A commonly suggested algorithm for this setting is "receding horizon control" (RHC), which computes the provisioning for the current time by optimizing over a window of predicted future loads. We show that RHC performs well in a homogeneous setting, in which all servers can serve all jobs equally well; however, we also prove that differences in propagation delays, servers, and electricity prices can cause RHC perform badly, So, we introduce variants of RHC that are guaranteed to perform as well in the face of such heterogeneity. These algorithms are then used to study the feasibility of powering a continent-wide set of data centers mostly by renewable sources, and to understand what portfolio of renewable energy is most effective.

Geographical Load Balancing for Online Service Applications in Distributed Datacenters

“…Most of the previous work [8,9,10,19] has focused on solving the GLB problem with continuous workload approximation. The problem can subsequently be solved using convex optimization methods.…”

“…Considering a constant communication delay for assigning a VM to a datacenter, a closed form solution can be found for (8) by using the M/M/1 queuing model, cf. [16].…”

“…These studies consider homogenous datacenters (where all servers are identical), which is far from the real-world situations. Reference [8] considers heterogeneous datacenters (comprised of servers with different performance and power dissipation figures, and even with instruction sets), which results in a more elaborate load balancing mechanism. Unfortunately, this work still ignores the heterogeneity of VMs, VM packing problem, and realistic VM migration cost and can result in low performance under realistic scenarios.…”

“…Most of the previous work that has focused on the GLB problem for online service applications, e.g., [8,9,10], simplify the VM placement and migration problem to a request forwarding problem for a VM type or a collection of VMs. This representation ignores the heterogeneity of VMs, the VM packing problem, and real VM migration cost and can thus result in low performance in a real cloud system.…”

Economics of Grids, Clouds, Systems, and Services

“…Lin et al [13] analyse a scenario with temporal variations in electricity prices and renewable energy availability for computation consolidation. Liu et al [15] define an algorithm for power demand shifting according to renewable power availability and cooling efficiency.…”