Virtual Machine Management: How to Balance Data center resource in practice

Today's demand for data centers is solid and effective, and this should be done first by "reasonable use". While idle resources are a waste of money for an environment. But if you fail to make a reasonable resource allocation to the frequently used data center, it will lead to a dangerous scene. A single hardware failure may cause other physical hosts to fail. The challenge for IT managers is to use and manage computing resources across the entire environment (usually including physical, virtual, and cloud resources). This article will discuss resource planning and problem mitigation as a means to achieve the goal of resource optimization. We will also explain how to solve the problem before it gets worse.

Best Practices for resource planning

Today, almost all data centers already have or will have some form of virtualized configuration. This requires more consideration in the virtualization of the physical platform. We have built a variety of virtual machine applications on a single hardware platform, computing resources sharing hardware platform CPU, memory, network resources sharing I sometimes traffic sharing disk. When it comes to resource management and problem positioning, proactive planning can help solve the problem of resource utilization.

Load Balancing of resources

Whether in a virtual or physical environment, it is important to understand which resources are being used and which virtual machines are assigned to them. For example, if you have a physical host on hand with a configuration of 8 cores and 24GB of random memory, engineers will not use all available resources, because then there is absolutely no room for adjustment and failure.

Figure 1 Citrix XenServer 6.0 Enterprise Management program configured as a single host with local memory support only. Deployments such as small businesses can be well run, but as the business changes, multiple hosts become necessary

As shown in Figure 1, a single XenServer host provides several high frequency workloads for end users. While this approach is feasible, there is little room for adjustment when facing the peak of business applications or the potential need to add virtual servers. In the example above, the engineer will have to remove the virtual resource from the other virtual machines--in this case, random memory--to achieve content growth on a single physical machine. Physical resources must be available for virtual machines as required. In any environment, there is a need to ensure that there is room to deal with emergencies and future expansions. This will be a great place for load balancing and virtual machine management.

Both virtual machines and physical machines must have appropriate types of resources assigned to them. When a frequently used workload is deployed, resources are planned and transferred without affecting other virtual or physical workloads. Also the case scenario, shown in Figure 2, we introduced a second physical host with similar hardware specifications to begin load balancing of virtual and physical resource units.

Figure 2 The Citrix XenServer 6.0 Enterprise Edition management program shows that multiple hosts are configured in the resource pool. The resource pool is capable of resource sharing because virtual machines can be implemented in real time between hosts within a storage area network (SAN)

In the new scenario, we have two physical hosts dedicated to the resource pool, where resources can be shared between virtual machines. Because each environment is unique, the load balancing process for physical resources needs to be individually planned. In this case, two physical hosts have additional resources available, as required by the current virtual machine. It also provides standby cpu,ram and storage for this specific environment to ensure the flexibility of virtual machine operations.

Similarly, many environments will consider load balancing for high availability (HA). With XenServer 6.0 as a demonstration management program, the built-in tools will help to complete the process. By using the "merge Server" high-availability feature, administrators can see which machines can safely fail over to another host. Here, engineers can determine how much resources each physical machine needs to handle its load. The most important thing to remember is that both servers meet the need for newly created virtual machines to have the resources available, failover and workflow automation to achieve load balancing. From the ha point of view, if one of the physical hosts mentioned above fails, another can host the key virtual machines on the failed host.

When resources are balanced between machines, virtual machines have the ability to move between physical hosts as needed without affecting the current resource state. Consider disaster preparedness (DR) as a possible example. If a physical host fails in this type of load-balancing scenario, the virtual opportunity is transferred to the next host that can find the available resources. If any one host is fully utilized, it is not possible to allow disaster preparedness and failover, simply because no resources are available on another host to support the incoming virtual machines.

Workflow Automation

Many environments that use both virtual and physical hosts may require workflow automation. For example. For example, Citrix Workflow Studio is a member of the Citrix IBuySpy Center product family. It is an IT process automation application that enables administrators to generate, dispatch, run, and manage workflows. These workflows bundle technical components together to implement the mechanization of repetitive configuration processes and help manage tasks to coordinate environment-triggered events. Based on the Microsoft.NET Framework, Windows Workflow Foundation and Windows Powershell,workflow Studio allow engineers to dynamically generate new virtual resources to respond to capacity requirements, Either planned or unplanned. In this scenario, the key point is to have the appropriate resources to correspond to the standby physical machine, so that the newly added virtual machine will have ram,cpus to use.

There is always a need to increase the virtual machine in an environment. The administrator decides how much resources to allocate for each new virtual machine. Too much or too little resource allocation can lead to waste of time and money. This is why it is important to develop a strategic plan for virtual machine management in an existing environment. Engineers can transfer workloads more efficiently by understanding what current data center available resources are.

This means that administrators need to be keenly aware of the physical and virtual environments they manage and how many users and machines can safely and efficiently handle tasks on that host. For example, the virtual desktop architecture. When a user logs on, it starts to consume a resource on a monitored machine, as shown in Figure 3.

Figure 3 The Citrix XenServer 6.0 Enterprise Management program shows that a stand-alone XenServer host is used only by the virtual desktop architecture. Virtual machines can be stored locally or on backbone storage LANs.

Currently, the machine shown in Figure 3 is not fully used. However, once the user flows in, the proportion of resources being occupied will grow rapidly, thus creating problems for this unbalanced environment. Use this data to create a virtual workload accordingly. For example, set a cap on resource usage so that in a properly managed datacenter, you can run a workload within the number of security values on a single physical machine.

Set resource reminders and alarms

Creating alerts and notifications in the data center can help maintain a healthy environment and improve virtual machine management. Finding problems before users perceive them, or before they compromise service level agreements (SLAs), can make the data center virtual and physical machines longer and more efficient. From a resource perspective, advanced management programs can set reminders and notifications, as shown in Figure 4.

Figure 4 Ctrixserver 6.0 Enterprise Management Program provides a "reminder" feature that can be configured on each virtual machine and physical machine. In this example, a reminder is set up for a Windows Server 2008r2 Enterprise Licensing server.

With alert monitoring, engineers can set up CPU, network, and disk alarms. Enable technicians to identify the problem of abuse early and take steps to mitigate resource allocation problems before impacting end users. In the planning and deployment process, setting up resource alarms is an important part. Many environments take this action to the last step, which can only cause the data center to fall into a resource-type problem quickly.

Using existing Third-party resource monitoring tools

Administrators often check for resources that directly affect individual physical or virtual hosts. In this case, there are some good third-party fine-grained monitoring tools that can help report special database servers, cloud machines, and other heavily used workloads. One of them was launched by uptime software company named up. Time tools allow administrators to monitor servers, virtual machines, clouds, and place, and other content. Use up. Time graphical server monitoring software enables an administrator to graphically and analyze all critical server resources running within the datacenter independently of any operating system. On the deep level, fine-grained monitoring of CPU, memory, disk, process, workload, network, user, service status and configuration data can help engineers rationally allocate and plan data center resources.

Another reliable network monitoring tool comes from the SolarWinds company. The tool, named Orion Receptacle configured Monitor (NPM), provides fine-grained network traffic and performance monitoring. To assist engineers in their daily work, NPM monitors can search for routers, switches, wireless access points, servers, and other SNMP-type devices, and analyze real-time, in-depth network performance statistics. For large data centers, NPM allows engineers to quickly browse core IT services and data center conditions by looking up filtered alarms that dynamically assemble related systems and devices.

In addition, resource issues can often be diagnosed and resolved using airborne tools. For example, Resource Monitor, launched by the Microsoft operating platform, can graphically illustrate resource utilization on a single machine, allowing administrators to understand how resources are being used.

Figure 5 The Resource Monitor shows the current memory usage on the Windows Server 2008 R2 Enterprise Exchange Server

In the Figure 5 example, the Store.exe of the server has a problem with RAM. Agents should understand that exchange software may be causing a ram-intensive phenomenon, so it is normal to see this usage. With this in mind, however, the engineer is able to choose either to increase the host resources or to transfer some of the load on the host to other machines.

Naturally, Resource Monitor has several settings to help engineers detect machines and find out which resources are being used. Another example is the amount of network processing. In Figure 6, we see a server in a normal operation. However, if a network spike is present, we can see the source and determine how the most effective mitigation can be made. To get a fine-grained panorama, engineers can drill down into an existing environment and build their own monitors to see where the data center is missing or where the data center is running.

Figure 6 Resource Monitor shows network traffic and usage on the Windows Server 2008 R2 Enterprise Exchange Server

Data Center Storage resource considerations

Storage resources can be both limited and expensive. Unreasonable storage utilization can cause operational problems and ultimately pay a high price in exchange for resolution. It is always important to monitor storage utilization in both virtual and physical environments. Intelligent storage tools reduce the pain of a huge workload by merging data and effective transport. One of the main vulnerabilities of data center storage LAN environment is the use of peak resources. For example, at a certain time, a large number of users access the system.

In the face of these situations, the disk becomes the main device used, and the operation is almost stopped. To cope with this situation, the storage LAN vendors are looking for solid technology and intelligent "re-recognition" caching mechanisms to mitigate performance bottlenecks.

The Figure 7 chart shows how the flash cache affects the NetApp 3000 series controller and its disk aggregation. This screenshot shows an activity with a total length of 80 minutes-the first 20 minutes of which is not cached.

Figure 7, a large number of users are accessing a workload that is heavily used. The device in this example is the NetApp controller. We can see the difference between the lack of cached disk performance and the organic load cache. These classes of solutions that help the data center run efficiently can make an environment more durable and smoother. In this case, there is no need to purchase a large amount of disk for resource distribution. Instead, they can leverage existing storage implementations to deliver a large workload more efficiently to end users.

Planning and attention pays off

Remember at any time that the computational resources are limited. The cost may be high when additional resources are needed to respond to unexpected events or shortages. This means that an administrator must have foresight in the overall data center environment and discover it early before the resource utilization issues affect the workload and end users.

(Responsible editor: Lu Guang)