America's smart PDU reveals the mystery of virtual power consumption

For most large companies or data center It executives, the energy problem is one of the key issues to be solved now and even in the future. No one can ignore this, because a lot of energy waste, will lead to limited funds can not promote it new technology innovation, electricity has become the whole operating costs of the enterprise to eat the "slot machine."

Many data centers, in order to reduce energy consumption and maximize the efficiency of the server, the introduction of http://www.aliyun.com/zixun/aggregation/13883.html "> Virtualization technology is unstoppable, but which servers can be virtualized?" In the same way, even if servers, networks, storage, etc. are applied, how much power savings do they bring? Rough computing and fine-grained management make the energy-consumption assessment of virtualization technology a blank for data center management.

The value of PDU needs further excavation

At present, enterprise users ' knowledge of PDU and IPDU is still mainly in the function of remote restart. This is because during the operation of various systems, there are often some unknown faults, especially when the physical server is loaded with multiple virtual machines, once the load is too large, restarting may be your last chance. Most of these hosts will not be in the local IDC room, the inevitable need for remote management. Through the use of PDU power sockets, especially IPDU power sockets, the administrator does not have to run to the scene to solve the host lock problem, easy to restart operation.

But very few people consider the other role of PDU, which has changed a lot in the cost management of data center, its core is in the management of power consumption. With PDU, IPDU and management software, users can clearly understand the power consumption of IT equipment. For example, the Power IQ management software of the U.S. national company it enables managers to improve the effectiveness of data center power supply through a single Web interface that clearly understands basic data such as power, current, voltage, temperature, and humidity of IT equipment, PDU exit groups, racks, and branch circuits. Strengthen capacity planning and improve the efficiency of energy use. These improvements can take advantage of the assessment before virtualization and, of course, the refined management of energy consumption after virtualization.

Server Power assessment helps virtualization process

Data center managers are always obsessed with how to determine which servers are not being used or underutilized. One of the methods they identified was to use CPU utilization as a yardstick to measure whether a server was fully utilized. However, this method does not always work. The server may sometimes seem busy, and it is in fact only doing level two or level three processing, not working on the server's most direct related tasks. Therefore, the assessment of energy consumption can effectively help users further advance the process of virtualization.

Currently, energy efficiency indicators for the measurement data center include: Power Usage effectiveness,pue and Datacenter Infrastructure Efficiency (data center infrastructure Efficiency,dcie). Both of these metrics take into account the energy consumed by the data center's power supply, cooling system, and it equipment, and are supported by the green grid. According to the Green grid study, the most effective way to measure the energy consumption of a data center It device is to measure the output of the PDU (Power distribution Units) in the computer room, which represents the total electricity that the data center transports to the server cabinet.

With the smart PDU and its management software, we can clearly see the efficiency of each server, because the most modest way to save energy is to shut down those idle or dormant servers that account for about 10% to 30% of the total data Center server. With the PDU's monitoring of energy consumption, we can have sufficient evidence to use virtualization to consolidate the underutilized servers and shut down unused servers.

Virtualization cannot be a "exonerated gold" in energy consumption

When it comes to green it, we can't avoid the two core objects of virtualization, energy consumption, but is the relationship just a simple combination? Although inefficient servers waste a staggering amount of energy, even if you deploy virtualization technology, there is a direct relationship between CPU utilization, the number of virtual machines on each physical host, and the amount of power consumed. Because the physical hosts that assume virtualization tend to improve their performance dramatically, it executives worry more about it. Because the new server CPU core increase, the physical host power consumption in the end increased how much, the entire room after the virtualization of the energy consumption reduced how much (this is worth affirming)?

Before, we could use the efficiency calculator introduced by the server vendor to make some assumptions about the power consumption, but these data are basic sources of energy, many of which run a specific workload without virtualization technology. When the number of virtual machines increases, the CPU of the server will run at full speed and the energy consumption will increase. At this point, the user can also use the virtual manufacturer to launch the calculation equation, but the conclusion and the actual running energy consumption results will be biased, because the server load is not fixed. Therefore, at any time to understand the power consumption of equipment changes, the development of targeted management strategy, as a virtual platform for the data center as important.

We analyzed the application environment of a financial enterprise, now they can use virtualization technology to achieve more than 400 low-end servers, and finally migrated to 85 new physical servers. This leads to a significant increase in server usage, but how much less energy is consumed after virtualization? Answer to us is: "A lot of reduction, but the specific data is not available." "Once there was data showing:" In Google's data center, each processing 100 searches, equivalent to 28 minutes of consumption of 60 watts of lighting. The three-day YouTube video stream consumes as much energy as it produces, protects, and delivers a DVD. "How do these data and comparisons get?" Believe that their work must be fine to the extreme.

So when it comes to energy saving, the 1th you need to know the precise power consumption data per host, if you do not know this, you can hardly find the opportunity to reduce energy consumption has been placed in front of you, as well as in the power consumption management to achieve refinement.

In the past, the data center power management is too extensive, the data center Pue is too high, because there is no detailed power data support, it is difficult to find the problem, not to mention the targeted solution. Then, after virtualization, real power consumption data has become the key to Pue computing accuracy, perhaps the smart PDU can help you complete this complex work. In fact, this is the work that PDU is doing.