Search

Uncorrectable memory errors are the leading causes of server failures in datacenters. Predicting uncorrectable errors (UEs) using the historical correctable error (CE) information helps for proactive replacement of memory hardware before the catastrophic events happen. In this paper, we perform an empirical study of UE prediction on the large-scale field data from more than 30,000 contemporary servers in Tencent datacenters over an 8-month period. We demonstrate that the traditional approach based on CE rate works poorly with a low precision. We then leverage the detail micro-level CE information to design several new predictors. The comparative study shows that the new predictor based on column fault identification boosts the baseline precision for a factor of more than 300% and at the same time also improve the baseline recall substantially.

The upper limb rehabilitation training is critical and indispensable for post-stroke patients. In recent years, an increasing number of rehabilitation robots have been well-designed, and the advantages of these robotic devices include making repetitive movements easily, automating the therapy process and supply therapists with a novel tool. Thus, we design a rehabilitation training and administration system for upper limb rehabilitation robot (TAS). In the system, the user information interfaces are introduced so that therapists can collect, store, retrieve and exchange patients' identification electronically. Moreover, the passive training method and virtual reality games are programmed to stimulate patients' enthusiasm in training process. Using the database, the basic information and historical training performance are stored for further analysis. Finally, the experiment demonstrates that our system is reliable and user-friendly to improve efficiency of rehabilitation therapies, which provide a feasible solution to the shortage of therapists.

In a typical data center, a significant portion of the total power consumption is spent on data center cooling. This paper addresses the problem of cooling inefficiency resulting from airflow mismatch between what is required by the IT equipment vs. what is supplied by the facility fans. The paper proposes two new virtual sensors for a server; volumetric airflow and outlet temperature sensor. A method for calculating server volumetric airflow in real time based on fan RPM values is outlined. The calculated volumetric airflow and outlet temperature are exposed to Intel Data center Manager (DCM) using IPMI commands. DCM can then aggregate server airflow to compute airflow demand by rack, row or data center level. The information can be used to control facility fans to improve cooling efficiency of the data center. This represents a significant improvement over use of expensive temperature & pressure sensors distributed throughout the data center to control facility fans.

The optimal technological requirements for flavonoids extracting by ultrasonic wave from rice plant (variety ASD7) and the insecticidal effect of the flavonoids on brown planthopper (BPH) were investigated. Four main factors, including ethanol concentration, ratio of liquid to solid, ultrasonic power and extracting time were firstly analyzed by singlefactor analysis, and followed by orthogonal experiments. The condition for extracting flavonoids from rice plant was optimized as following: 70% ethanol, a solid-liquid ratio of 1:25, a power of 250 w, and an extracting time of 50 min. Under these conditions the extraction rates of 5 rice varieties was 0.34% (TN1), 0.38% (Rathu Heenati), 0.36% (ASD7), 0.33% (Mudgo), and 0.28% (Xiushui 11), respectively. The survival rates of BPH fed on artificial diets with different dosages of flavonoids were observed at 24 h and 48 h. When treatment with 0.1 mg/ml extracting flavonoids, the survival rates of BPH were 34.0% (24 h) and 6.0% (48 h), significantly differed from the control (78.0%, 24 h; 32.0%, 48 h). The survival rate of BPH decreased with the increase of the flavonoids concentration, especially, at a concentration of 0.5 mg/ml of extracting flavonoids, no BPH survived for more than 24 h. The results indicated that flavonoids had an insecticidal effect on BPH.

Over the last two decades, artificial neural networks (ANN) have been applied to solve a variety of problems such as pattern classification and function approximation. In many applications, it is desirable to extract knowledge from trained neural networks for the users to gain a better understanding of the network's solution. In this paper, we apply REFANN (rule extraction from function approximating neural networks) in dividend events study. Based on our study of 1530 dividend initiations and 692 resumptions events from April 1965 to December 2000, we find that the positive relation between the short-term price reaction and the ratio of annualized dividend amount to stock price is primarily limited to 96 firms that have high dividend ratio and small firm size. The results suggest that the degree of short-term stock price underreaction to dividend events may not be as dramatic as previously believed. The results also show that the relations between the stock price response and firm size is also different across different types of firms. It is suggested that drawing the conclusions from the whole dividend events data may leave some important information unexamined. Our rule extraction method may shed some lights on further empirical research in corporate events studies because more information can be drawn from the data.

We propose a neural network based saliency analysis to identify important information related to investment decisions. We apply our approach to 753 US stocks and 1960 head-and-shoulders patterns. The results show that among all variables that are related to technical patterns, membership value and pattern length as well as firm size have the highest saliency coefficients and are the most relevant to future returns. The results are consistent with previous studies and the practice of technical analysis. Our study suggests that saliency analysis offers much greater advantage when compared with correlation coefficients in revealing the relationships in the stock market. Average investors can use saliency analysis to sort through information and find the more important information to focus on, therefore shortening their learning curve of becoming expert investors.