Xiaonan Zhao

A high-level energy consumption model for heterogeneous data centers

Abstract Data centers consume anywhere between 1.7% and 2.2% of the United States’ power. A handful of studies focused on ways of predicting power consumption of computing platforms based on performance events counters. Most of existing power-consumption models retrieve performance counters from hardware, which offer accurate measurement of energy dissipation. Although these models were verified on several machines with specific CPU chips, it is difficult to deploy these models into data centers equipped by heterogeneous computing platforms. While models based on resource utilization via OS monitoring tools can be used in heterogeneous data centers, most of these models were linear model. In this paper, we analyze the accuracy of linear models with the SPECpower benchmark results, which is a widely adopted benchmark to evaluate the power and performance characteristics of servers. There are 392 published results until October 2012; these servers represent most servers in heterogeneous data centers. We use R-squared, RMSE (Root Mean Square Error) and average error to validate the accuracy of the linear model. The results show that not all servers fit the linear model very well. 6.5% of R-squared values are less than 0.95, which means linear regression does not fit the data very well. 12.5% of RMSE values are greater than 20, which means there is still big difference between modeled and real power consumption. We extend the linear model to high degree polynomial models. We found the cubic polynomial model can get better results than the linear model. We also apply the linear model and the cubic model to estimate real-time energy consumption on two different servers. The results show that linear model can get accurate prediction value when server energy consumption swing in a small range. The cubic model can get better results for servers with small and wide range.

Key Technologies for Green Data Center

With the repaid development of multi-media and network, information are generated and spread quickly than before. The industry is expected to meet the needs by producing high performance CPU and high-density disk. But dramatic increase in energy requirements becomes a new challenging. Energy and human cost become the major parts of data center. More and more vendor provides efficiency products, such as MAID(Massive array idle disk). SPEC has developed a power benchmark for single server (SPECPowerssj2008), SNIA has developed a green storage measure standard. In this paper, we summary several technologies to reduce energy consume of data center. These technologies can be divided into three levels: power and refrigeration, green components, intelligence energy management. At the last part of paper, we introduced the test standard of energy usage of server and storage.

A Hierarchical Storage Strategy Based on Block-Level Data Valuation

Hierarchical storage is the core for information lifecycle management (ILM), and data classification based on its value is the key issue for hierarchical storage. In this paper, a block-level valuation model is proposed to classify data in disk. The model parameters is determined by comparing the factors in the view of block-level with in the file-level view data value, and it also discussed how to gather and measure the parameters in the model from a real storage environment. At last the model is validated through a real case study. The experiment shows that the model is feasible and valuable for data classification in storage subsystem, even it knows nothing about the data context in the disk.

Block-Level Data Migration in Tiered Storage System

Managing data growth continues to be the biggest challenge for IT. And tiered storage, a solid way to Information Lifecycle Management (ILM), is very helpful to reduce the cost and improve the storage efficiency in data center. However, data migration is the primary challenge to implement tiered storage management, and its hard to be resolved by traditional data migration solutions, which are mainly relying on experience of administrators and manual migration according to storage capacity usage and data access frequency. This paper proposes a novel bi-directional migration policy (migration with double thresholds based on feedback, MDTF), based on block-level data valuation and fully automation process. MDTF aims to achieve balance between storage QoS and migration costs by double thresholds, to narrow the migration scope of block-level data objects, and the analysis shows that MDTF is an efficient block-level data migration policy comparing with traditional migration policies.

A scheme to ensure data security of cloud storage

Cloud storage offers great convenience to users. However, the data security concerns may impede cloud storage wide adoption. Although there are many schemes to solve the problem, they mostly focus on an aspect of data security. They have not a detailed discussion about the data security in cloud. In this paper, we proposed approach provides a scheme to allow users to check the integrity of their data in the cloud. We also present the performance criteria about the scheme, which can help the researchers to optimize their mechanism efficiently and effectively. The scheme is lightweight, efficient and robust.

Storage System Optimization in Data-Intensive Environment

With the repaid development of multi-media and network, information are generated and spread quickly than before. New application such as search engine, video share depends on the scale and access speed of huge data set. Data-intensive computing became a hot point of computer science. In Data-intensive environments, data access ability is the bottleneck instead of computing ability.In this paper, we present a framework to manage data distribute and recent research results to accelerate data access in data-intensive environment.

An Improved Approach on B Tree Management for NAND Flash-Memory Storage Systems

With the significant growth of the markets for consumer electronics and various embedded systems, flash memory is now an economic solution for storage systems design. Tree index structures have been adopted over flash-memory and the system performance can be significantly improved.But with the very distinctive characteristics of flash-memory, the overhead of intensive byte-wise operations are caused by record inserting, record deleting, and tree reorganizing. Such actions result in a large number of data copyings (i.e., the copying of unchanged data and tree pointers in related nodes). In this paper, we introduced segment, segment summary and segment Mapping, and they can reduce the possibility of related node update problem andimprove the system performance.

Storage Performance Optimization Based on ARIMA

This paper analyzes the potential causes of the performance bottleneck in I/O access paths of storage architecture and proposes a predictive approach based on feedforward to optimize the I/O performance of storage subsystems effectively, which uses a time series analysis method based on ARIMA to build the predictive and monitor model of the performance. This approach can improve the availability of the storage subsystem effectively and decrease TCO by decreasing the possibility of I/O bottlenecks.

The Cost Performance of Hyper-Threading Technology in the Cloud Computing Systems

Hyper-Threading (or HT, for short) is a technology used in some Intel CPUs. Intel claims that it can use processor resources more efficiently. Many past studies have evaluated the performance of the technology in HPC clusters. In this paper, we discuss the advantages and disadvantages of Hyper-Threading using in the cloud computing systems. We evaluate the performance and energy cost of Intel CPU with Hyper-Threading enabled and disabled on virtualization environment. Our results show that Hyper-Threading technology can get better performance in most cases on a physical machine. The performance of a single core in a virtual machine is slightly lower when HT is enabled. But it doubles the number of available cores.

A Framework to Measure Storage Utilization in Cloud Storage Systems

Cloud storage systems aim to offer cost-effective storage services. The key is sharing resources between multiple users by virtualization technologies. Storage resources in cloud systems can not be reclaimed even when users do not access their data for a long time. Storage resources must be shared through space sharing rather than time sharing. Existing technologies improve storage utilization at various layers and data sets, making it difficult to analyze the efficiency of a cloud storage in a holistic way. To address this problem, we propose an evaluation framework to study the impacts of a wide variety of I/O techniques on an enterprise-scale cloud storage. The framework offers storage utilization evaluation from both the users and the vendors’ perspective.