Peafowl

Abstract

The traffic load sent to key-value (KV) stores varies over long timescales of hours to short timescales of a few microseconds. Long-term variations present the opportunity to save power during low or medium periods of utilization. Several techniques exist to save power in servers, including feedback-based controllers that right-size the number of allocated CPU cores, dynamic voltage and frequency scaling (DVFS), and c-state (idle-state) mechanisms. In this paper, we demonstrate that existing power saving techniques are not effective for KV stores. This is because the high rate of traffic even under low load prevents the system from entering low power states for extended periods of time. To achieve power savings, we must unbalance the load among the CPU cores so that some of them can enter low power states during periods of low load. We accomplish this by introducing the notion of in-application CPU scheduling. Instead of relying on the kernel to schedule threads, we pin threads to bypass the kernel CPU scheduler and then perform the scheduling within the KV store application. Our design, Peafowl, is a KV store that features an in-application CPU scheduler that monitors the load to learn the workload characteristics and then scales the number of active CPU cores when the load drops, leading to notable power savings during low or medium periods of utilization. Our experiments demonstrate that Peafowl uses up to 40-54% lower power than state of the art approaches such as Rubik and DPM.

A Study of Sleep Time Alignment of CPU Cores for Power Saving

“…Peafowl [5] saves power by scaling in the threads that process requests in accordance with the amount of workload, thereby reducing the number of active CPU cores during periods of low load. However, the application must be modified to scale in and out of threads, which does not meet requirement (b).…”

Hector: A Framework to Design and Evaluate Scheduling Strategies in Persistent Key-Value Stores

“…C3 also maintains the current count c of remote pending requests for each replica, and an history of observed latencies R, which are finally used to compute a score according to the cubic function R − μ−1 + μ−1 (1 + cm + q) 3 , where R, μ−1 and q are respectively the EWMAs of observed latencies, service times, and queue sizes, while m is the number of servers. Here, using a cubic term aims to penalize servers with longer queues, which is expected to lead to better balancing.…”

“…Other systems leverage scheduling strategies to improve other metrics such as power consumption. Peafowl [3] unbalances the load of processing local requests on a storage node in a key-value store, to allow some of the cores to enter a low-power state, thereby reducing the energy footprint of the node in periods of lower activity. Multiget APIs allow clients to request multiple values in a single query.…”

Use Two Kind Hybridization of the Chaotic Peafowl Algorithm with the Hummingbird Algorithm

“…Where males are distinguished by their long tail compared to females with short or no tail, and their behavior is divided into two parts: social behavior and spatial behavior. [7][8][9][10].…”