Virtualization Data Center performance reliability Diagnostics and configuration

Tenkine Server Channel March 5 message The performance diagnostics on the physical server are relatively straightforward before virtualization. In most cases, administrators can http://www.aliyun.com/zixun/aggregation/7432.html > fix problems by reconfiguring, upgrading, or patching up the server. However, virtualization to performance Diagnostics adds a new, complex virtualization layer. Running multiple workloads is certainly beneficial for improving server utilization, but hardware resource sharing can sometimes lead to unpredictable results that adversely affect the performance of the workload in unexpected ways.

Assume that a database server shares the same host server as a multimedia server. The server has enough resources to handle the resource requirements of the workload, and in normal usage mode, both workloads deliver good enough performance. Now that users are complaining about poor database query performance, IT pros often check the database server virtual machine to respond to users ' complaints, hoping to find that the configuration has changed or there is insufficient resources. But after checking the database server, the only problem seems to be that local disk IO is too frequent when database query performance is low, but local disk activity is not related to the database virtual machine. However, another technician found that the multimedia server virtual machine handles a large number of video processing requests when the database response is slow.

Virtualization can complicate the tracing of the source of the problem because it may not be possible to adjust the resources of a virtual machine or migrate a virtual machine without considering other load conditions on a particular system. To diagnose problems more efficiently, IT professionals need to use business service management tools that identify the physical path of the virtual machine and the applications that each virtual machine is running.

In addition, improving the reliability of the virtual server can start with a careful examination of the server hardware and features. Let's start with a server that has a redundant power supply. Enterprise Server will include two redundant power modules, any one power module can support the normal operation of the server. When a power supply module fails, the backup power module will take over without affecting the system's normal operation. Redundant power modules are certainly not a new idea, but it is important to ensure that redundant power supplies replace the old single power system in the future.

Virtual machines run in server memory, so IT professionals can increase server availability by selecting systems with memory reliability technology. Other reliability techniques typically include configuring multiple network adapters for the server. When all hardware is working properly, more ports provide additional bandwidth and ensure connectivity to all server workloads. If a NIC port fails, the server can still provide network connectivity, minimizing the impact on the virtual machine.

But IT pros are not just buying more reliable machines-it's important to take the appropriate policies and measures to solve the server failure when there are errors. High availability technology does not ensure that there are no errors on the server-these techniques only help the server continue to run when there is an error. When the memory module or power module fails, the reliability of the server will be reduced until the technician completes the repair, so consider the alarm, failover process, and the virtual server troubleshooting requirements.

One of the most effective ways for servers to achieve reliability for critical business is to use server clusters. Clusters create server groups, and each server in the cluster runs redundant virtual machines. When a server fails, the cluster software removes the failed server from the cluster, and other replicas of the virtual machine take over the business without causing a service outage. The primary advantage of clustering is that servers in a cluster often give up many high-availability features-business is often taken over by other servers in the cluster.

In short, virtualization makes data center operational dimensions a new situation, although virtualization improves server efficiency. But when a problem arises, the diagnostics become complicated. So redundancy on hardware and software becomes very important when virtualization is enabled in the datacenter. Second, management tools, how to implement the server cluster monitoring and diagnosis is also critical.