Server virtualization is in full swing and unstoppable. There are also many legends about the power savings brought about by server virtualization. Everyone does not doubt the positive contribution of server virtualization to green environmental protection, but how to quantify the power savings brought by server virtualization? This article attempts to lead everyone to make some useful explorations, hoping to help us understand server virtualization more clearly. The importance and urgency of a green data center.
Of course, today's virtualization products gradually increased, but up to now we are still widely spread since last year's VMware Virtual Infrastructure 3 (Virtual Infrastructure 3, referred to VI3). So we're here to explore this and compare the power expenditures before and after deploying VMware VI3, so that we can clearly understand how the emerging technology of server virtualization can change the energy consumption of our data center.
For the convenience of explanation, a user is assumed here. Let us call this user Shanghai Oak Company. Let us study the power saving situation after the deployment of virtualization by Shanghai Oak.
Shanghai Oaks company's IT profile
because our main server virtualization has brought energy savings, therefore, we assume that Shanghai Oak's PC servers as follows:
Number of single CPU servers
300Number of dual CPU servers
500
>The number of four CPU servers
200
Server integration is the most common application scenario for server virtualization, which is based on the server integration model. Power savings from virtualization. First look at the power consumption of the data center. According to the actual situation, the power consumption of the data center can be divided into two categories:
· Power consumption of computing architecture (IT load): server hardware, network switching equipment, storage network equipment, etc.;
· Power consumption of network critical physical architecture (referred to as: NCPI, non-IT load): transformers, UPS power supplies, power cables, fans, air conditioners, humidifiers, lighting, pumps, etc. To complete the entire study data center power consumption, it is necessary to consider each part of the above relates to the direct power
Data Center consumption
. Since we only study server virtualization here, we only study the power consumption of server hardware, combined with virtualization, to study the difference in server power consumption before and after virtualization.
To calculate the power consumption of the entire data center server, you only need to superimpose the power ratings of all servers in the data center. Since the superposition of rated power is the maximum power that the server may consume, in order to reflect the actual energy consumption, it is necessary to multiply the corresponding adjustment factor to reflect the server power consumption in the steady state of the data center. The steady state coefficient is an empirical value. According to the experience of American Power Conversion Corporation, "...the nominal power of most IT equipment is probably more than 33% of the actual energy consumption."
Server power consumption is usually calculated (eg: 1U, 2U, 4U, etc.) based on the height dimension of the server, and that the CPU of the machine height dimension we have here the following relationship:
· 1 Ke CPU = 1U
· 2 Ke CPU = 2U
· 4 Ke CPU = 4U
· 8 Ke CPU = 6U
· 16 Ke CPU = 12U
· 32 CPU = 24U
'500')this.width='500';if(this.height>'400')this.height='400';" Border=0>
Shanghai Oaks' overall electricity bill savings:
'500')this.width='500';if( This.height>'400')this.height='400';" border=0>
The above calculation is based on the following assumptions:
After deploying VMware VI3, As CPU utilization is greatly increased, the power consumption of the server will increase accordingly. In other words, the power consumption of a single server after virtualization is higher than before virtualization. The exact relationship between increased CPU utilization and increased power consumption is difficult to quantify and varies from CPU to CPU and from server model. However, through Dell's online tools, power consumption can be estimated based on the type of load. Dell offers the following types of load: idle, average load, I/O intensive, and compute intensive. A computationally intensive load may consume 150 W more power than an idle state and may consume 80 W more than the average load. (Note: This tool does not provide actual utilization of each load CPU).
Here we Shanghai Oaks company as an example, to calculate electricity costs. According to Shanghai's electricity price in 2006, we assume that Shanghai Oak Company's electricity cost per kilowatt is 0.8 yuan, then the power consumption of Shanghai Oak's server hardware is:
'500')this. Width='500';if(this.height>'400')this.height='400';" border=0>
Therefore, the annual electricity bill before deploying the VMware virtualization solution The expenses are:
'500')this.width='500';if(this.height>'400')this.height='400';" border=0>
After deploying the VMware virtualization solution, the annual electricity bill is:
'500')this.width='500';if (this.height>'400')this.height='400';" border=0>
By deploying VMware virtualization solutions, the annual electricity bill for direct data consumption of data center servers is:
'500')this.width='500';if(this.height>'400')this.height='400';" border=0>
the power consumption of the data center cooling systems are all electronic devices generate heat
data center, including servers, switches, network storage devices and the like disposed IT Also includes a power supply system equipment, such as UPS, power distribution system. Of course, air conditioning equipment, lighting and even people will generate heat. As before, for the sake of discussion, only the heat generated by the server hardware is discussed here.
data center design largely determines the thermal efficiency and the cooling cost, many data centers still use a ventilation mode by the front-to-back, all servers in one direction, which will result in a server from behind The hot air that is emitted will go directly to the front of another server. A better design idea is the thermal-channel/cold-channel design layout, which reduces the temperature steps from the front-to-back ventilation mode.
Even with an optimized design, approximately 25% of the redundant air flow is required. When the computer room air conditioning and refrigeration system fails, redundant airflow ensures data center cooling needs. Moreover, many data centers have hot spots, and the heat density of hotspots tends to be much higher than other areas, and the concentrated redundant airflow can meet the president's cooling needs of the data center.
In addition to requiring airflow redundancy, the data center requires additional airflow to meet cooling needs, as sometimes moisture can cause cooling efficiencies. Of course, proper humidity is also required. Excessive drying may cause electrostatic discharge, which may cause damage to certain equipment. However, most air conditioning systems condense water vapor to reduce ambient humidity. In order to maintain proper humidity, a humidifier may be required.
Generally, most of electric power consumed in the entire data center into heat, and therefore, server hardware rated power output substantially corresponds to the thermal energy. In addition, according to the results of the Hewlett-Packard laboratory, each watt of heat dissipation requires 0.8 watts of electrical energy for cooling. (We define this as the load parameter L). Based on the above data, the cooling energy consumption of the data center can be calculated as follows:
'500')this.width='500';if(this.height>'400')this .height='400';" border=0>
Savings in electricity bills for Shanghai Oaks cooling:
'500')this.width ='500';if(this.height>'400')this.height='400';" border=0>
The above calculation is based on the following Assumptions:
The calculations here do not include the cost of installment payments and the maintenance costs of the power cooling system. If you need to take into account these costs, please calculate separately according to your actual situation.
Let's be calculated based on the company's case oak Shanghai. By adding the server's rated power, it can be calculated that the stable power consumption is 407 kW before the deployment of virtualization, and the power consumption is 51.6 kW after deploying the VMware virtualization solution. This is the heat dissipated by the Shanghai Oaks data center server. Therefore, the cost of cooling these heat can be calculated as follows.
Before deploying a VMware virtualization solution, the cooling cost of the data center server is:
'500')this.width='500';if(this.height >'400')this.height='400';" border=0>
After deploying the VMware virtualization solution, the cooling cost of the data center server is reduced to:
< Br>
'500')this.width='500';if(this.height>'400')this.height='400';" border=0>
Deploying VMware virtualization solutions, data center server cooling power consumption annual electricity savings is:
'500')this.width='500';if(this.height>'400 ')this.height='400';" border=0>
Conclusion: Server virtualization brings huge electricity savings to the data center
As can be seen from the above calculations, data center server virtualization can bring a lot of power savings, and the electricity cost savings are considerable. By deploying VMware virtualization solutions, Shanghai Oaks can save a total of electricity bills each year. RMB 5,254,151.00 yuan, which is incredible before we calculate it. Of course, the cost savings may vary depending on the specific case, but the savings in electricity bills are significant, which is just one of the good benefits of deploying server virtualization.
Therefore, deployment Server virtualization has become a top priority for large data centers and one of the keys to effectively controlling power shortages in data centers.