Per-Server Dominant-Share Fairness (PS-DSF): A multi-resource fair allocation mechanism for heterogeneous servers

Abstract

Users of cloud computing platforms pose different types of demands for multiple resources on servers (physical or virtual machines). Besides differences in their resource capacities, servers may be additionally heterogeneous in their ability to service users -certain users' tasks may only be serviced by a subset of the servers. We identify important shortcomings in existing multi-resource fair allocation mechanisms -Dominant Resource Fairness (DRF) and its follow up work -when used in such environments. We develop a new fair allocation mechanism called Per-Server Dominant-Share Fairness (PS-DSF) which we show offers all desirable sharing properties that DRF is able to offer in the case of a single "resource pool" (i.e., if the resources of all servers were pooled together into one hypothetical server). We evaluate the performance of PS-DSF through simulations. Our evaluation shows the enhanced efficiency of PS-DSF compared to the existing allocation mechanisms. We argue how our proposed allocation mechanism is applicable in cloud computing networks and especially large scale data-centers.

An Efficient and Fair Multi-Resource Allocation Mechanism for Heterogeneous Servers

“…In case of heterogeneous servers with placement constraints, TSF is shown to satisfy Pareto optimality, envy freeness and sharing incentive properties [11]. However, we show by example that this mechanism may not satisfy bottleneck fairness (neither in the sense of Definition 2, nor in the conventional sense based on considering a single virtual server introduced above [12]).…”

“…This approach is amenable to a distributed implementation. It results in an enhanced performance over the existing mechanisms, and satisfies certain properties essential for fair allocation of resources [12].…”

“…In this paper, we build upon and generalize our proposed PS-DSF allocation mechanism [12] to capture the trade-off between efficiency and fairness. We concisely summarize our contributions.…”

“…According to Definition 2, RAM is identified as the bottleneck resource in this example. To allocate the RAM resources in a fair manner, each user should be allocated x 1 = x 1,1 = 2, x 2 = x 2,1 = 6, x 3 = x 3,2 = 8 and x 4 = x 4,2 = 8 tasks, respectively (This allocation results from our proposed PS-DSF allocation mechanism [12]). On the other hand, the DRFH mechanism would instead identify network bandwidth as the dominant resource for the first two users and RAM as the dominant resource for the last two users.…”

“…In [12] we proposed a multi-resource fair allocation mechanism, called PS-DSF, which is applicable to heterogeneous servers in the presence of placement constraints. The intuition behind PS-DSF is to capture the impact of server heterogeneity by measuring the total allocated resources to each user explicitly from the perspective of each server.…”