HP DL360 Power basics for IT professionals - Page 25

percent. When a server is con d for N+1 redundancy, N that is

Get HP DL360 - ProLiant - G3 PDF manuals and user guides
UPC - 613326948835
View all HP DL360 manuals
Add to My Manuals
Save this manual to your list of manuals

Page 25 highlights

3,000 watts per rack. As of 2005, it appears that data centers averaged eight to twelve servers (2U to 3U each) per rack and wattage of 5 to 6 kilowatts per rack. It is important to remember that with the increase in density 1U and server blade racks can draw two to three times that amount, or 10 to 18 kilowatts of power. Growth in data center power requirements is averaging 20 to 30 percent per server generation. Replacing just one older server with a new one could increase power density by 90 to 140 percent. For the immediate future, it would be sensible to anticipate power densities of more than 1 kilowatt per U and 25 to 30 kilowatts per rack. To anticipate maximum real-world power loads and capacities, HP highly recommends that customers use the HP Power Calculator software. Power Calculators are based on actual system measurements, so they rely on rational factors to calculate the upper range of power needs (which can vary widely). Power Calculators significantly and intelligently streamline planning for current and future capacity as well as redundancy. Power Calculator results are based on actual system measurements, taken on live systems running Microsoft® Windows NT® while all major components (central processor, memory, and drives) are being exercised at 100 percent of their duty cycle. Therefore, calculations may be higher than the actual power draw of a customer's configuration. HP Power Calculators are accessible at the following http://h30099.www3.hp.com/configurator/powercalcs.asp and www.hp.com/go/bladesystem/powercalculator. Another essential server strategy is to plan for redundancy. When a server is configured for 1+1 redundancy, both power supplies are powering the server. This shared-load approach ensures that the redundant power supply is able to assume the full load and limits the step-load inrush of current to 50 percent. When a server is configured for N+1 redundancy, N (that is, 2, 3, or possibly more) power supplies are powering the server while one additional power supply stands ready to support the load in case of a power supply failure. As the number of servers and blade enclosures in a rack increase, the UPS type must change to anticipate load growth. Best practices include deployment for redundancy by using 80 percent of capacity, careful load planning, and effective coordination between facilities, IT, and operations personnel. Traditional data centers are designed for five- to ten-year life cycles. Within a five-year growth cycle, the power consumption and server density could more than double, making it imperative that power, cooling, and redundancy be continually reviewed along with compute and storage capacity. 25

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34
  • 35
  • 36
  • 37
  • 38
  • 39
  • 40
  • 41
  • 42
  • 43
  • 44
3,000 watts per rack. As of 2005, it appears that data centers averaged eight to twelve servers (2U
to 3U each) per rack and wattage of 5 to 6 kilowatts per rack. It is important to remember that with
the increase in density 1U and server blade racks can draw two to three times that amount, or 10 to
18 kilowatts of power.
Growth in data center power requirements is averaging 20 to 30 percent per server generation.
Replacing just one older server with a new one could increase power density by 90 to 140 percent.
For the immediate future, it would be sensible to anticipate power densities of more than 1 kilowatt
per U and 25 to 30 kilowatts per rack.
To anticipate maximum real-world power loads and capacities, HP highly recommends that customers
use the HP Power Calculator software. Power Calculators are based on actual system measurements,
so they rely on rational factors to calculate the upper range of power needs (which can vary widely).
Power Calculators significantly and intelligently streamline planning for current and future capacity as
well as redundancy. Power Calculator results are based on actual system measurements, taken on live
systems running Microsoft® Windows NT® while all major components (central processor, memory,
and drives) are being exercised at 100 percent of their duty cycle. Therefore, calculations may be
higher than the actual power draw of a customer’s configuration.
HP Power Calculators are accessible at the following
and
www.hp.com/go/bladesystem/powercalculator
.
Another essential server strategy is to plan for redundancy. When a server is configured for 1+1
redundancy, both power supplies are powering the server. This shared-load approach ensures that
the redundant power supply is able to assume the full load and limits the step-load inrush of current to
50 percent. When a server is configured for N+1 redundancy, N (that is, 2, 3, or possibly more)
power supplies are powering the server while one additional power supply stands ready to support
the load in case of a power supply failure.
As the number of servers and blade enclosures in a rack increase, the UPS type must change to
anticipate load growth. Best practices include deployment for redundancy by using 80 percent of
capacity, careful load planning, and effective coordination between facilities, IT, and operations
personnel. Traditional data centers are designed for five- to ten-year life cycles. Within a five-year
growth cycle, the power consumption and server density could more than double, making it
imperative that power, cooling, and redundancy be continually reviewed along with compute and
storage capacity.
25