Green Data Centers:
Introduction:
Data center is a warehouse where we can store, manipulate, manage and disseminate the data. All the data handling equipment consume a huge amount of energy. To run a data center, it needs power or energy i.e. electricity. It has some effects on surroundingswhich can be either environmental for example CO2 emission, heat etc. or non-environmental decrease in efficiency. In order to make it efficient with both aspects environmental and non environmental there are some mechanisms so that a data center could be energy efficient and environment friendly. We can call it green data centers.
Digital data grows tremendously day by day. Growth of data demands of data center for storage and management purposes. According to a report, digital data will increase approximately 44 times of current situation, which increases demand of data centers. Power consumption by data centers will increase accordingly,which causes different problems in addition with CO2 emission. Another report says that CO2 emission from data centers may exceed from any air line CO2 emission by 2020.which is an alarming situation for data centers. So, there is a need of such data centers which are energy efficient as well as environment friendly to meet the needs of present situation. Data centers which will not update their status with the need of time have risk to pay more than double for the same amount of energy requirements. Therefore, there is a need for making data centers green so that they can be energy efficient as well as environment responsible by paying appropriate cost.
Problem Statement:
Tremendous growth of data results in very high power consumption and demands power management. High power consumption results in high cost, excessive CO2 emission and many more problems.
Hypothesis:
By upgrading and improving the metrics and technologies we can make a data center cost and energy efficient.
Research Questions:
1. How to make a data center which consumes less power as well as it is environment friendly?
2. What are the metrics which should consider for green data center?
3. Design such algorithms which distribute the workload among multiple data centers so energy efficiency can be obtained?
Literature review:
For energy efficiency and cost efficiency in traditional data centers some metrics and technologies are devised so that a data center can become a green data center.
Lizhe Samee described some categories of metrics which we can be used to make data center green e.g. its CO2 emission, humidity level, cooling system or thermal control, power or energy usage as well as multiple indicators at different levels in a data center can be used as a metric. In developing the metrics to make a data center green, some driving features should be considered. First of all, the metric which is under consideration should be economical. Secondly, it should be environment friendly. Moreover, there should be a performance metric which can keep an eye on whether energy efficiency by greening a data center put any negative impact on the computation performance or not. Performance metrics which should be taken under consideration are like itshould be stable enough so that different types of data centers can adapt it. It should be balanced so that whole system can perform perfectly . For all types of situations which can be expected in the future, metric should be evaluated before so that it can cater a foreseen problem [1].
Xiao Yi discuss in their study that, to make data centers green, it has to use green cooling systems, green components etc.By making a group of different metrics technologies are developed. These technologies are divided into three levels. These levels are Energy efficiency and cooling, energy efficient components, energy efficient management[2]. Energy Efficiency and Cooling is hardware or system based level. Electricity and refrigeration in a system use a major fraction of total cost used to run a data center. According to a survey report less than 45% of total cost used in IT equipment and most of the remaining part used for cooling and energy efficiency. In a report about energy consumption says that 49% of total power used in non-computing areas like power distribution to different components, to run UPS, generators and fans etc.
There are many different cooling techniques like “Dynamic Smart Cooling” it use sensors to limit heating effects, by using this technique usage of power decreases from 117kw to 72 kw.Second technique is ”Air Conditioned System” works on per rack level and uses air and ice water to cool the system but this system has a drawback of excessive noise . Third technique is “Spray Cooling System” it cools the system on chip level.it is 3500 times better than the technique which uses air to refrigerate the system. The disadvantage of this system is its complexity. It is hard to design and implement.
Second major level to make a green data center is having energy efficient storage components in the system. In past, hard disk drive is used as a secondary storage device which consumes high power, gives low reliability and has less speed. To cater all these problems Solid State Disk is used today. SSD gives high latency rate, high reliability, consumes less power and with very small power consumption about.2.5 kwh its capacity is up to 2TBs but with all these advantages it has some problems which are dominant on its advantages like gives bit errors, has limited number of erases, low random write performance, high price etc.
Third and last level is energy efficient data management Softwares e.g. “Storage Virtualization”, “Deduplication”, “Thin Provisioning” and “Massive Array of Idle Disk”are used to manage data. In all of the softwares management is done by solving problems like under utilization, over allocation, redundancy, power consumption by idle disks etc.
Kianiand Ansari (2015) studied the green energy and devised a mechanism that helps in maximizing its use to cut electricity costs especially in computing. These authors suggested maximum greener workload and minimum brown workload. This comparison will allow having maximized green work load and minimized brown workload. The reason is that the demand of servers is increasing because every day huge data is generated and there should be servers to keep that data. These servers run on high computation powers consuming electricity [3]. The authors introduced a model called GOOD or Green and economic workload Distribution which integrates multiple data centers which consumer green energy.
Critical analysis:
Some metrics discussed here are although enough to standardize a data center with energy efficiency but these metrics cannot ensure the performance of data centers in terms of sustainability, reliability and connectivity etc. There should be some metrics which can make standards for the performance efficiency. Some technologies are developed to make system efficient in physical manner like sprayers and sensors are used for cooling, SSDs are used as storage components but they are expensive, noisy and complex. These techniques should be easy to use, cost effective as well as environment friendly and consistency is an important requirement for storage devices to get reliability. Some management techniques are used to manage resources and requirements between different clients. It reduces power consumption and storage requirement in an efficient way but still need improvements with the demand of time.
Conclusion:
For greening a data center there should be some mechanisms through which it becomes more energy efficient then legacy data centers without putting any negative impact on environment. In this paper some metrics are discussed which are used to standardize a data center. Some techniques are studied to make a data center green and review an algorithm GOOD to distribute workload among a group of data centers so it can perform in energy efficient way with an advantage of cost efficiency.
References:
[1] Review of performance metrics for green data centers:a taxonomy study by Lizhe Wang • Samee U. Khan
[2] Key Technologies for Green Data Center by Xiao Zhang, Xiao-nan Zhao, Yi Lin, Lei-jieZeng
[3]Kiani, A., and Ansari, N., (2015), “Toward Low-Cost Workload Distribution for Integrated Green Data Centers”, IEEE COMMUNICATIONS LETTERS, VOL. 19, NO. 1