Virtual combat: One of host design sizing

In the early stages of a virtualization project, estimate the budget to have a general plan for the hosts you need. What type of host is required? What is the configuration for each host? How much do you need?

Let's assume a scenario to start the discussion.

A company has 50 physical servers and intends to migrate to the virtualization platform. Here are three steps to help you find the answer.

Analysis of requirements to collect data

Be sure to have a good understanding of the needs. Do you need to migrate existing physical hosts to this virtualized environment? What is the application running on the physical host? Average and peak value of resource usage?

VMware Capacityplanner can easily collect this information from existing physical hosts and generate very intuitive reports. Or you can use tools such as Perfmon to collect data manually.

Assume the following scenario:

The average core number of 50 physical hosts is 4, and the average CPU frequency is 2000Mhz. The CPU occupancy rate at peak is 12%

See more highlights of this column: http://www.bianceng.cnhttp://www.bianceng.cn/Servers/virtualization/

We need to pay special attention to, we use the peak occupancy rate, that is, the maximum load situation of the demand for computational resources. For some applications to know what things, what time period is likely to be peak, data acquisition in the comprehensive.

The choice of two virtualization hosts

Host already available

It may be that the user has cooperated with a vendor, and it is best to assess whether the vendor has a suitable host. Specific reference VMware HCL (Hardware Compatibility List)

Scale out vs Scale up

Scale up refers to the use of a high configuration host, thereby reducing the number of hosts. Scale out refers to the use of relatively low configuration of the host, a large number.

There are a lot of things to consider in this trade-off. In the case of permitting, I am personally inclined to low configuration, a large number.

Advantages:

• The number of virtual machines running on each host is low, and HA can quickly reboot the affected virtual machines on other hosts in the event of a host outage. Thus minimizing the impact on the user.

· DRS can better balance load across multiple hosts

• There is a lot of room for a host upgrade to vmotion a virtual machine on that host in a short time.

Possible disadvantages:

• The data center space that you occupy may be large. So if the number of hosts is large, the blade server may be a better choice.

• Possible support costs will be high. In some server-supported outsourcing contracts, support costs are associated with the number of hosts