A cyber-physical integrated system for application performance and energy management in data centers

Abstract

Both performance and energy cost are important concerns for current data center operators. Traditionally, however, IT and mechanical engineers have separately optimized the cyber vs. physical aspects of data center operations. In contrast, the work presented in this paper jointly considers both the IT -cyber -and the physical systems in data centers, the eventual goal being to develop performance and power management techniques that holistically operate to control the entire complex of data center installations. Toward this end, we propose a balance of payments model for holistic power and performance management. As an example of coordinated data center management system, the energy-aware cyber-physical system (EaCPS) uses an application controller on the cyber side to guarantee application performance, and on the physical side, it utilizes electric current-aware capacity management (CACM) to smartly place executables to reduce the energy consumption of each chassis present in a data center rack. A web application, representative of a multi-tier web site, is used to evaluate the performance of the controller on the cyber side, the CACM control on the physical side, and of the holistic EaCPS methods in a mid-size, instrumented data center. Results indicate that coordinated EaCPS outperforms the cyber and physical control modules working separately.

Bridging Physical and Digital Traffic System Simulations with the Gulliver Test-Bed

“…Cyber-physical systems target varying areas from water distribution modeling [16], to medical applications [12], and data center performance and energy management [6], to name a few. Several works have studied digital vehicular systems by considering digital road traffic and network simulators [1,2,8,21].…”

Demand side management with consumer clusters in cyber‐physical smart distribution system considering price‐based and reward‐based scheduling programs

“…Similar toresidential loads, the cyber and physical aspects of industrial processes, datacentre operation performance and management were also considered for DR programs. [28]. Thecyber‐physical infrastructure and computer‐aided operation of overall smart gridparadigm including smart buildings and smart cities can reduce overall energyconsumption [29, 30].…”

Evaluating the need for complexity in energy-aware management for cloud platforms

“…As a result, management tasks aimed at optimizing metrics involving multiple parameters, such as balancing CPU usage and temperature distribution, must understand and consider this complexity. For instance, a management method aimed at maintaining performance and balanced power utilization, cannot simply look at the nodes' CPU, memory and other resources' usage, but instead must consider the design of the specific hardware platform where such management is conducted [5,4]. Or, management methods concerned with balancing CPU usage and avoiding hotspot creation, must be aware of the infrastructure topology and physical properties of the environment.…”

“…On top of this data storage layer, we run map-reduce Hadoop jobs to aggregate the per-VM or per-set of VMs energy usage over time, to gain insights into certain datacenter or system behaviors, and to validate models that can be used at runtime to manage such systems [4,5].…”

“…This model, further adopted by Cloud Computing, is based on the observation that typical resource utilization levels are low -in few 10s of percent, however hardware is not "power proportional", thereby utilizing significant power/energy levels in spite of the relatively low consumption of IT resources. In this context, researchers in industry and academia are actively developing sophisticated solutions to manage how consolidated cloud workloads use the underlying cloud platform, so as to ensure increased energy efficiency, while still satisfying application/client performance demands [11,2,13,16,18,17,15,8,5,6,1].…”