Research On Green Data Management Of Cloud Storage System

With the increasing popularity of data-intensive applications and services,large-scale data centers consumed enormous power resource. The energyconsumption of storage system accounting for25%to35%of data centers energyconsumption. Cloud Storage System has been widely used. Therefore, how to reducecloud storage devices energy consumption in large data center is an urgent problemneed to be solved.This paper aimed to combine data classification, data placement, data replicationand dynamic voltage management technology, through ratio nal and efficient datamanagement in ensuring the quality of service requirements to reduce energyconsumption maximum extent possible. This paper introduces the background, thedevelopment status and the related academic work of saving technology in stora gesystem, and describes the significance of cloud storage saving technology. Amongthem, this paper emphatically introduces the neural network model, the graycorrelation analysis technology and the technology of GridSim used in simulationexperiments. In order to solve the problem of energy consumption in the cloudstorage system, the main contribution of this paper is purpos ing green dataclassification strategy based on anticipation (AGDC) and data classification-basedgreen gear-shifting mechanism (DGLG), the main work of these two contributions issummarized as follows:(1)This paper proposes a green data classification strategy based on anticipation(AGDC), which classify the data in cloud storage system: the hot data is stored in thehot disk region; the cold data is stored in the cold disk region. AGDC employneural-network prediction on seasonal data, predicting the temperature of data in thenext period, executing seasonal data migration in cold and hot regions. This paperalso adopts a new correlating algorithm on new data, analyzing its relations with olddata in the storage system and predicting the data temperature. New energyconsumption model is also established in this paper. Simulation experiments basedon GridSim show that the cloud storage system with green data classificationstrategy based on anticipation has a good effect on reducing energy consumption. Atthe expense of average response time of0.005s, our algorithm saved about16%of energy consumption when compared TDCS.(2)Based on anticipation of the green data classification strategy (AGDC), thispaper purposed data classification-based green gear-shifting mechanism (DGLG):The frame designed a new data partitioning strategy, data replication managementstrategy and proposed energy gear-shifting mechanism based on the data partitionand data replication management.1) Data partitioning strategy through anticipationof the green data classification strategy (AGDC) classifies the data into cold data,hot data, seasonal data, and new data and put it into the appropriate zone. The hotdata and seasonal data which pre-divided for hot individually placed in source hotzone of hot zone; the cold data and seasonal data which pre-divided for coldindividually placed in source cold zone of cold zone (Definitions related to see4.1).2) Data replication management strategy depend on the nature of the data developsthe appropriate number of replica and replica placement. The more hot data containsmore backups than cold data, the new data and seasonal data which pre-divided forhot only have min backups in hot zone.3) Based on the above data partitioning, datareplication management, this paper proposes energy gear-shifting mechanism thatautomatically gear-shifting through neural network model to predict follow-upperiod assignments. Experiments based on GridSim show that: the energyconsumption of gear-shifting mechanism is cost effective. It can achievebetter energy efficiency in the storage system with the small file, which saved about43%average energy at the expense of about1.6ms average response time and themaximum energy savings is about78%, minimum energy savings is about16%.DGLG lifting gear mechanism which purposed based on AGDC combines dataclassification, data placement, data backup and dynamic voltage managementtechnology to reach the green data management in the cloud storage system.