How the cloud computing environment measures server efficiency

From an infrastructure perspective, an efficient cloud computing environment must also be an efficient server. And the efficiency of the server is mainly embodied in two aspects, on the one hand, the efficiency of the server itself, that is, utilization and execution efficiency.

Server utilization is critical

Improve the utilization of the server, generally is the main embodiment of efficiency. The reason is simple: If a server, the performance is good, its utilization is low, then the efficiency of this server is very low. From a cost perspective, we also want the server to work at full load state, not always in idle state.

So how to improve the utilization of the server? We need to first analyze the reasons for the low utilization of the server. The first is that there are not so many workloads-this involves effective investment and knowledge of the server. Then, if the workload is too high, it will also result in the calculation of I/O congestion, resulting in low utilization. The above two questions are actually a task scheduling and load balancing problem. On the other hand, utilization is also related to server configuration bottlenecks.

In some of the earlier machines, although the server computing performance is very strong, the core number is also many. But utilization is not always there because memory is too small to create a system bottleneck-now, in front of the Intel Xeon E5, AMD Bulldozer architecture, and Power7, the memory capacity is no longer a problem. And new bottlenecks are more frequent--the frequency sensitive application is not suitable in the multi-core environment, but needs similar Power7 such high frequency server to raise the utilization rate. Conversely, some thread-sensitive applications require a multi core process such as the AMD Bulldozer architecture. Intel's CPU architecture has a better performance at the HPC level.

It can be seen that the understanding of their application types and reasonable configuration of the server is to solve the system bottleneck, improve server utilization of the important factors.

What about load Balancing? First of all, the traditional method of load balancing is the real-time monitoring of the server, with manual or load balancing algorithm to split the workload, the advantage is to achieve a simpler, the disadvantage is time-consuming and laborious, always to adjust. There are also manufacturers to introduce self-learning and a number of strategies for load balancing software, and even hardware products to complete job scheduling. But this is a palliative, not a permanent solution.

So the virtualization of this unification means has become the universal application of the means, and rise to become the cornerstone of cloud computing. The principle is very simple, that is, pooling resources-using the virtual method of the server cluster computing capacity, storage capacity, network bandwidth virtual into a resource pool, in which the dynamic allocation of computing resources, so there is no simple server partition, also bypassed the problem of load balancing.

However, let's refocus the topic on the server level--where is the direct link between utilization and virtualization? Is that some of the previously available computing resources can be created and distributed to other application tasks, and this is separated from the existing applications.

This sentence is very awkward, you can think of it is to split yourself in half, a dry heavy physical life, the rest of the strength to do a little easy work, the overall goal is to take your strength full-this is to improve utilization.

To sum up, we draw two conclusions: first we need to select the appropriate server according to the type of application, so as to maximize the utilization rate; Then on this basis, if the virtualization solution is adopted, more resources can be extracted from the server and the utilization rate will be increased on a larger scale. It also provides a good foundation for cloud computing.

Energy efficiency has many considerations

If the server with high utilization efficiency is high, then energy consumption is another factor. Today's users are very objective, the server considerations are not only performance, utilization. In addition to a one-time procurement costs, energy efficiency is also the focus of research direction, so the large amount of it funds will be spent on maintenance and electricity and so on.

So what kind of server is energy efficient? You have to know that the biggest power consumption in the server is the processor, so choose the new architecture, the new process of the CPU platform, the most energy-efficient server has been selected. This is a simple principle: The new process allows the CPU to improve performance and save power, while some of the new CPU features, such as Intel's power Gate,turboboost, and so on, are better balanced in terms of consumption and CPU performance.

In addition, many servers are designed with energy efficiency in mind. For example, Dell's latest 12g series server, HP's GEN8 series of products have adopted the 80plus Platinum level standard power supply, the power conversion efficiency of more than 90%, compared with the past to save a lot of power loss. and low voltage memory and intelligent control of the fan will further reduce the power consumption of the server.

This is also true of management, with Dell's new IDRAC7 Management suite on the 12g server, which enables very intuitive monitoring and observation of server power consumption, and a combination of Intel NodeManager management technology for server-powered control. HP uses a new Lifecycle lifecycle management tool in the GEN8 series of servers, even monitoring and controlling power consumption through the ipad.

So, servers that make cloud computing more efficient are actually in many ways consistent with the ideas that cloud computing advocates. For example, energy saving, high utilization, delicate control, dynamic task load distribution and so on. The choice of efficient servers, in turn, can help push cloud computing environment more efficient, the real meaning of complementarity.

(Responsible editor: The good of the Legacy)