How to get from virtualization to cloud data center

The operation of modern enterprises is inseparable from data. The data center hosts thousands of servers to support applications, provide information, and automate multiple processes. There is no sign of a weakening in the demand for data center capacity. Thousands of of servers running together will consume a lot of electricity, resulting in increased operating costs, resulting in a growing global greenhouse gas.

Our research shows that through tighter management, companies can increase the energy efficiency of data centers by one fold, thereby reducing costs and reducing greenhouse gas emissions. In particular, companies need to be more aggressive in managing technical assets and increasing the level of utilization of existing servers; companies also need to more accurately predict the driving effects of business requirements on the capacity of applications, servers, and data center facilities to control unnecessary capital and operating expenses.

The efficiency of data centers is a strategic issue

The amount of money spent on building and operating data centers is rising in the company's IT budget, resulting in fewer budgets for urgently needed technology projects. The data center construction plan is a board-level decision. At the same time, regulators and external stakeholders are increasingly concerned about how companies manage their carbon footprint. Adopting best practices not only helps companies to reduce pollution, but also enhances their image as good corporate citizens. A costly problem

Today, the company's analysis is increasingly complex, customers require real-time access to the account, and staff are also looking for new technology-intensive collaboration methods. Thus, even as the economy slows, demand for computing, storage, and network capacity continues to grow. To address this trend, the IT department is increasing its computing resources. In the United States, the number of servers in the data center is increasing by about 10% a year. At the same time, in emerging markets such as China and India, where institutions are becoming more complex, more operations are being automated, and more and more outsourced data services are being done here, the number of data centers is growing faster. This demand for computational resources cannot be suppressed, leading to a steady rise in global data center capacity. For now, this growth has not shown signs of coming to an end, usually in a period of economic recession, when it will only go into moderate growth.

This growth has led to a surge in it costs. Data center spending typically accounts for 25% of the total enterprise IT budget if you calculate the cost of facilities, storage devices, servers, and people. As the number of servers continues to grow, electricity prices are rising at a faster rate than revenues and other it costs, which will only increase. Every year, the cost of running these facilities is rising by as much as 20%, while total it expenditure is growing at a rate of only 6%.

The increasing expenditure of data centers has changed the economic structure of many enterprises, especially information-intensive enterprises such as finance, information services, media and telecommunications companies. Over the past 5 years, the investment required to set up a large enterprise data center has risen from $150 million to $500 million. In it-intensive enterprises, the cost of the largest facilities is approaching $1 billion trillion. This spending has squeezed the capital for new product development, reducing the economic benefits of some data-intensive products and reducing profits. In addition, rising energy consumption produces more and wider carbon footprint, resulting in environmental degradation. For most service industries, data centers are the main source of greenhouse gas emissions for businesses. Over 2000-2006 years, the power to store and process data doubled, with the average electricity consumption of each data facility equivalent to the sum of 25,000 households. There are 44 million servers in the world, consuming 0.5% of total power. Today, data centers ' carbon emissions are close to those of countries such as Argentina and the Netherlands. In the US alone, the data center's projected electricity growth by 2010 is equivalent to generating 10 new power plants. Current projections suggest that, if demand is not curbed, global data centers will have 4 times times as much carbon emissions in 2020.

Regulators have noted these trends and are urging companies to come up with a solution. The US Environmental Protection Agency (EPA) recommends that, as a first step in establishing operational efficiency standards, large data centers should use energy meters. At the same time, the EU has issued a voluntary code of conduct, which describes best practices for running data centres with higher energy efficiency. As data centers continue to rise, the government may be putting more pressure on reducing emissions.

Far-reaching challenges

In information-intensive organizations, many departments and levels of people can make decisions that affect the efficiency of data center operations. Financial traders can choose to run complex Monte Carlo (Monte) analysis, and drug researchers can decide how many clinical experimental image data to store. Managers who are responsible for application development can decide how much programming work is needed to meet these needs. The management of the server infrastructure can make equipment procurement decisions. The facility supervisor can determine the location of the data center, the power supply, and the timetable for installing the equipment before the projected demand arises.

These decisions are usually made in isolation. The Sales Manager may choose to change the transaction from overnight settlement to instant settlement, and financial analysts may want to store several copies of historical data, and they do not consider what impact this will have on the cost of the datacenter. It is very rare for application developers to optimize their work to minimize server usage and to develop design applications that can be shared across servers. The manager who buys the server may choose the product with the lowest price or the one they are most familiar with. But these servers may waste data center power or space. Often, managers buy more equipment to ensure that they have enough capacity in the most extreme cases, which can cause excess capacity. Managers tend to build facilities with excess space and high refrigeration capacity to meet demand in extreme situations or to respond to emergency expansions.

These decisions add up throughout the organization and will have a significant impact on costs and the environment. In many cases, companies can deactivate existing parts of the server without reducing their data management capabilities, and shelve plans to buy new servers. This can be achieved with the help of some well-known technologies. Virtualization, for example, actually runs the application by looking for the spare part of the server to achieve capacity-sharing purposes. But companies do not necessarily do that because no senior executive can assume the "end to end" responsibility. Within the organization, managers act in the best interests of their own, which results in inefficiencies in most data centers, often running only one software application per server.

We analyzed nearly 500 servers in a media company, where utilization was less than 3% of One-third, while less than 10% accounted for two-thirds. Although there are many out-of-the-box management tools for tracking usage, the company does not use any of them. Globally, we estimate that the daily utilization rate of servers is generally only 5% to 10%, which creates a waste of energy and money. In response, data center managers generally answer that these servers are designed to provide capacity in extreme cases, such as coping with the shopping spree the day before Christmas. But in general, this assertion is not tenable, because the data show: if the average utilization rate is extremely low, then the utilization rate of peak period will be very low. In addition, the number of data facilities continues to climb, but the storage of servers and related equipment sometimes accounts for only half of the capacity of the data facilities, indicating that billions of dollars in capital expenditure have been wasted. Even if the company reports that the data center is full, but walks along the data center aisle, it often finds that there are lots of empty slots on the server racks, and the devices that were originally placed in those slots have been eliminated.

This inconsistency occurs in part because of the difficulty of predicting data center requirements. The time frame for operations is a problem. The design and construction of the data center typically takes 2 years or more, and the expected service life is at least 12, so capacity is set before the business unit generates actual demand. At the same time, there is a lack of understanding of how business decisions interact with each other, how they are translated into requirements for new applications, and how much server capacity is needed to meet demand. For example, if customer demand increases by 50%, it is difficult for many companies to predict whether the capacity of servers and data centers will need to increase by 25% or increase by 100%. In extreme cases, we found that some of the facilities were in mid-air for years after they were put into operation, while others soon found it necessary to build a new one after they had built a data center.