What kind of server makes the cloud more efficient (1)

In terms of infrastructure, an efficient cloud computing environment must also have an efficient server. The efficiency of the server is mainly reflected in two aspects: the efficiency of the server itself, that is, the utilization and execution efficiency. On the other hand, it is the server's energy efficiency, that is, the energy consumption. This is two important indicators for measuring server efficiency, and we can have a lot of interesting discussions on these two aspects.

Server utilization is critical

Improving server utilization is generally the main embodiment of productivity. The reason is simple: if a server has good performance and its utilization is low, the efficiency of the server will be very low. In terms of cost, we also hope that the server is working at full capacity, rather than always idle.

So how can we improve the server utilization? We need to analyze the cause of low server utilization. First, there are not so many workloads-This involves effective investment and understanding of servers. Then, if the workload is too high, it will also cause computing I/O congestion, resulting in a low utilization rate. The above two problems are actually task scheduling and load balancing issues. On the other hand, the utilization rate is also related to the configuration bottleneck of the server.

In some earlier machines, although the server's computing performance was very powerful, there were also a large number of cores. However, the overall utilization rate does not go up because the memory size is too small, which forms a System Bottleneck. Now, in the face of Intel Xeon E5, AMD bulldozer architecture, and minicomputers represented by Power7, the memory capacity is no longer a problem. New bottlenecks emerge even more-frequency-sensitive applications are not suitable for multi-core environments, but require high-frequency servers like Power7 to improve utilization. In turn, some thread-sensitive applications require multi-core processing such as AMD bulldozer architecture. At the HPC level, Intel's CPU architecture has a better performance.

It can be seen that understanding the application type and configuring the Server properly is an important factor to solve the System Bottleneck and improve the server utilization.

What about Server Load balancer? First, let's talk about the traditional practice. Traditionally, Server Load balancer uses real-time monitoring of servers and uses manual or Server Load balancer algorithms to share the workload. The advantage of this is that the implementation is relatively simple, the disadvantage is that it is time-consuming and labor-consuming and always requires adjustment. Some vendors also launched self-learning and load balancing software with numerous policies, and even hardware products to schedule jobs. However, this is, after all, a way to cure the problem.

As a result, virtualization has become a universal means of application and has become the cornerstone of cloud computing. The principle is very simple, that is, pooling resources-virtualize the computing capabilities, storage capabilities, and network bandwidth of Server clusters into resource pools using virtualization techniques, and dynamically allocate computing resources among them, in this way, the Server Load balancer problem is bypassed without simply separating servers.

However, let's refocus on the server layer-where is the direct connection between higher utilization and virtualization? In the past, some idle computing resources can be virtualized and allocated to other application tasks, which are separated from existing applications.

You can think of splitting yourself into half, one dry-weight physical activity, and the rest of your work can be done easily, the overall goal is to take up your strength-this is to improve utilization.

To sum up, we come to two conclusions: first, we need to select an appropriate server based on the application type, so as to maximize utilization, and then if we adopt a virtualization solution on this basis, in this way, server resources can be extracted to a larger extent to improve utilization. It also laid the foundation for cloud computing.