Integrated Wiring data center integrated design technology (1)

In the face of future needs, data center design becomes increasingly difficult. As a concept, the overall design is designed to meet the needs of data centers by evaluating the needs, availability requirements, funding capabilities, and various business models of each customer.

In a recent survey, the respondents were asked to state their company's minimum requirements for normal operation time. About 25% of the respondents answered this question. Their minimum requirement is 99.999%! It is not difficult to see why most enterprise data centers regard system availability as the most important design factor. At the same time, it is difficult for practitioners and designers to determine the "ideal design" for each company ". Many reasons are involved in preparation. The primary cause is the long service life of the data center. A data center may last for 10 years or even longer, and designers usually do not intervene for such a long period of time to obtain the feedback needed to determine how to improve the design.

At the same time, factors such as minor design defects and poor operating procedures will also cause many problems. In the end, it may be because people do not have enough knowledge about this that makes it difficult to improve the design. We can only learn and apply the accumulated experience from building a data center to a new data center and improve it.

Throughout the existing high-availability data centers, the design and operation modes of each data center vary greatly, but there are many similarities in many aspects. By integrating the data, feedback, and goals of various data centers, we hope to facilitate and assist in the design or improvement of data centers.

The main problems to be discussed in this article are:

◆ Unbalanced supply and demand of power supply design capacity;

◆ Predict future power supply demands;

◆ Quickly change the power density of the data center;

◆ Establish infrastructure that can effectively allocate power;

◆ Add hot restrictions on new data centers;

◆ Create and optimize power redundancy options;

◆ Understand the impact of operations and training;

◆ Explain the feasibility of "99.999%;

◆ How System Recovery affects availability.

UPS power supply design capacity cannot be balanced with actual needs

When designing and constructing a data center, people regard the scale of the predicted data center as a basic program. However, if the predicted design life of the data center is 10 ~ 15 years later, this task becomes very difficult. During this life cycle, devices in the room may be refreshed or changed four or five times by a new generation of devices. The rapid update of IT equipment makes the most basic future capacity forecast a bubble.

For example, Moore's Law points out that the integration of Integrated Circuits doubles every 18 months. For each generation of products, the additional transistor needs to increase the power supply capacity proportionally. Correspondingly, the microprocessor's heat output also increases significantly. Intel's latest Pentium 4 chip generates approximately million heat. Compared with 486, the latter has less than 10 million heat. The increase in chip density increases the generation of heat at the same rate. As we can see from the data center design, the power density requirements have rapidly expanded over the past few years. Not long ago, 50 W per square meter was a typical value, but most designers have increased this value to 150 ~ per square meter ~ Million or more. Over the past few years, the system infrastructure has been dramatically expanded to meet the ever-changing demands for power capacity and cooling in data centers.

Predicting power supply in the future requires this challenging task to lead to one of the most serious data center problems, namely the imbalance between supply and demand of UPS design capacity. The emergence of this problem not only affects the efficiency, utilization and support capabilities of the data center, but also hinders the optimal use of capital. For example, after a large data center is built, 2 ~ 4 years can be achieved or close to the design capability. Therefore, there was little demand for power capacity in operations in the past few years. At this time, users often spend millions of dollars to build a data center that can achieve all the designed capacity, including purchasing Several megawatt redundant power systems to support the data center. The system can use all capacity only after a minimum of several years of low-capacity operation. In this case, a large amount of money is used for initial capacity construction to meet the needs that will arise after many years. It is ironic that while using all the power capacity, the data center still has a large footprint for use. This is because the power density of the device has been increasing for two years, and the power supply capacity has been fully used, but the footprint is very adequate. The ever-changing equipment technology makes the situation worse. Typically, a data center requires an overhaul every four years. In addition to overhaul costs, the risk of data center downtime is also greatly increased during the implementation process.

Typical utilization curve of large Internet data centers

This is especially dangerous for device hosting and internet service providers that use rental data centers as their primary revenue sources. These companies have to pay for the entire facility before the formal operation, but the income corresponding to this cost is often received after many years. For such capital-intensive industries, it is extremely dangerous to bear a huge cost load before obtaining income (1 ). It should be noted that the utilization curve may change significantly, and two years is usually a very optimistic figure. It can be seen that the balance between supply and demand almost never exists, which will lead to huge costs and fail to achieve optimal utilization.

Curve of power supply and demand in a scalable system

Shows the power supply and demand relationship of the design scheme with higher scalability.

In, the system design allows capacity upgrades. Supply always exceeds demand. However, this special design method also has limitations. Infrastructure changes pose risks to data center operations, which are rare in four phases. In fact, after the initial construction of the data center is completed, the entire project usually needs to be completed within one or two phases.

In order to improve the imbalance between power supply design capacity and actual capacity, it is important to predict the future needs first. Despite the difficulties and unsatisfactory results, it is crucial for long-term success. Secondly, the design process needs to integrate the flexibility and scalability required by the previous design. Third, to design a data center, you must pay attention to the customization of solutions to make them more standard. This not only reduces design costs, but also accelerates project implementation.