Search

Large data centers consume huge electricity every year. The power constraint is key factor for data center operator to increase rack server density, meanwhile a lot of provisioned power is reserved as the safe margin to handle the rarely peak power demand or power source failure, leading to considerable power under-utilization as well as implicated performance loss comparing to design target. This paper studied the problem of rack power utilization from internet data centers and analyzed approaches to optimize rack power utilization for the rack equipped with rack wide battery backup system. The dynamic rack power provision (DRPP) is proposed to address rack power under-utilization problem without expanding rack power capacity and an autonomous power control algorithm is discussed. By using power capacity locked in battery backup system to handle the sporadic power spike, the dynamic rack power provision is able to provision more servers into the rack by keeping no change of rack power budget. At last, this paper discussed the evaluation methodology and presented the experimental results, covering density improvement, performance impacts, and battery capability. The results showed 15%–25% utilization improvement for the experimental cloud workloads with the method proposed by this paper.

This paper introduced one optimized proactive cooling management approach based on power variation trend analysis. Through analyzing the data center historical power telemetries, the power predictor is able to predicate power variation with 5– 15 minutes granularity. The cooling controller uses the observed heat information and estimated thermal variation trend to drive CRAC to manage temperature situation at prediction window. To validate cooling results from different cooling parameters, one risk level evaluation method is proposed and the experiments for different prediction window are conducted and the result is presented.

A novel polarization insensitive metasurface for monostatic Radar Cross Section (RCS) reduction is designed, fabricated, and investigated. The proposed metasurface is comprised of carefully arranged unit cells with spatially varied dimension, which enables approximate uniform diffusion of incoming electromagnetic energy and avoids reflection at the specular direction. A theoretical model based on reflectarray theory is introduced to demonstrate the operating principle of the metasurface. By utilizing Gravitational Search Algorithm (GSA) and a fast far-field scattering pattern analysis method based on Inverse Fast Fourier transform technology (IFFT), a rapid and efficient design method is demonstrated. A metasurface sample is fabricated and tested to validate RCS reduction behavior predicted by full wave simulation. An over 20dB RCS reduction is observed in the vicinity of 9.5GHz, indicating our metasurface may be potentially applied in future stealth technology.

In today's data center, in order to achieve higher density and better energy efficiency, the traditional server systems with dedicated fans and power supply units are evolving to the Rack Scale System with shared cooling and shared power supply at rack level, some examples are Open Compute Project and China Scorpio Rack. This evolution brings the challenges of how to efficiently control the rack cooling and effectively manage the rack power at coarse granularity level, e.g. at the whole rack or half rack level, while no increasing in system cost and management complexity. In this paper, the management architecture for the Rack Scale System is introduced, the cooling zone, power zone, and the parameters for thermal and power management are elaborated. Some new virtual sensors for thermal management are also introduced in this paper, the modeling method for the volumetric airflow virtual sensor is described, and the validation test results are presented. With the rack thermal and power management, the IT devices can be closely associated with the datacenter facilities, and thus be intelligently orchestrated and adapted to the workload for TCO (Total Cost of Ownership) saving and reliability improvement.

Rapid growth of internet services and mobile devices has led to more and larger cloud data centers. The hyper scale cloud data center consumes enormous amount of electricity and cause pressure to operation cost and infrastructure management. The industry has made great progress in improving power usage effectiveness through innovation and infrastructure upgrade. Recent research is focusing on dynamically adjusting workload placement according to realtime power and thermal telemetry of datacenter infrastructure to reduce the pressure to datacenter power and thermal as well as to improve datacenter power usage effectiveness (PUE), like thermal awareness scheduler (TAS). To achieve higher density and longer lifecycle of high value computing component, the conventional rack-mount server system is evolving to rack scale server system with power and cooling units moving to rack level to share with multi server systems, like Facebook-led Open Compute project and Project Scorpio developed by Chinese internet giants – Baidu, Alibaba and Tencent. The traditional thermal awareness workload placement assumes all server systems within clusters are uniform with discrete power units and cooling units, and there are no power and thermal correlation between different server systems. However, with power and cooling units moving to rack level, the power and thermal correlation between different server systems must be considered while calculating the optimal workload placement. To address these challenges, in this paper, we propose a framework of multi-tiers thermalintelligent workload placement and corresponding thermal management algorithms accustomed to rack scale server systems with shared power and cooling units. This paper evaluated these thermal management algorithms from performance, benefits, as well as their usage scenarios. The prototype and experiment introduced in this paper run over OpenStack managed cluster, but the thermal-intelligent workload placement and corresponding thermal management policies introduced by this paper aim to provide one common framework in addition to Cloud OS, like Big data software stacks, even customer's distributed computing and storage systems.

In this work, an ACP was prepared and applied in the packaging of the RFID tag inlays through flip-chip bonding process. The shear strength, the radio frequency performances (Resonance Frequency and Scattering parameters-S11) before and after the aging tests (high humidity and temperature test, 85 °C, RH 85%) and the Maximum Read Distance of the prepared RFID tag inlays were used to evaluate their reliabilities. And the results were compared with those of RFID tag inlays packaged with a commercial ACP under the same bonding conditions. It was found that the radio frequency performance of the High Frequency RFID tag inlays packaged with the ACP prepared here and the commercial ACP, almost were identical and remained unchanged after the aging test. And the shear strengths of them both increased after the aging test. It was revealed that the ACP prepared here has the same performance of the commercial product. The prepared ACP maybe performed better than the commercial product from the relationships between the maximum read distances of the UtraHigh Frequency RFID tag inlays and the aging hours. It can be concluded that the preapared ACP can be applied in the RFID tag inlays assembling industry. It is another choice for the RFID tag manufactures. And this test system can also be used to evaluate the reliability of the RFID tag inlays.