CONSIDERATIONS FOR A HIGHLY

AVAILABLE INTELLIGENT
RACK POWER DISTRIBUTION UNIT
A White Paper on Availability
 Introduction
Data centers are currently undergoing a period of great change. Data center managers are struggling to keep pace with
growing capacity needs while working under the constraints of tightened budgets and energy efficiency initiatives and the
challenges presented by new technologies such as virtualization and cloud computing are transforming.
As data center environments become more dynamic and complex, many organizations are taking a more proactive approach
to management and gaining better control of their data center operations so they can maintain or improve availability in
increasingly dense computing environments while reducing costs and boosting efficiency. One area of improvement is inside
the rack with the growing importance of intelligent rack Power Distribution Units (rack PDUs).
As the last link in the power chain delivering critical power to IT loads, intelligent rack PDUs are a strategic asset for achieving
high availability through elevated levels of responsiveness to change in data center capacities and densities. The emergence
of data center infrastructure management (DCIM) is further increasing the role of the intelligent rack PDU within the data
center. Data center managers are taking advantage of the benefits provided by the technology, including access to rack-level
and IT equipment power consumption, visibility into rack-level environmental conditions, the ability to directly control power to
IT equipment and rack-level capacity and power management.
This white paper discusses the considerations that need to be made when investing in intelligent rack PDUs to ensure that
they do indeed provide a high availability solution.
Five aspects of a high availability design for Intelligent Rack PDUs include:
1. Reliability

2. Functionality

3. Fault Tolerance

4. Maintainability

5. Adaptability.

2
Reliability
Intelligent rack PDUs (Figure 1), offering flexible and
comprehensive remote management and real-time
monitoring capabilities, can provide the best view of IT
power consumption and rack operating conditions.
However, while these new rack PDUs offer advanced
capabilities, they are not much use if their core function is
compromised or impeded: providing basic power
distribution in any circumstance. There are a number of
factors that should be considered to help ensure the
primary function of intelligent rack PDUs.

Figure 2. Located at the back of the rack toward the hot-aisle, rack
PDUs can be exposed to temperatures exceeding 50 degrees Celsius.

Low Idle PDU Power Consumption

As organizations invest in higher density IT equipment, the
number of components contained inside a rack PDU has
increased in order to provide higher levels of intelligence.
However, the overall form factor has remained relatively
Figure 1. Today’s intelligent and adaptive rack PDUs provide value small. It is important to realize that higher power
beyond just power distribution, enhancing business agility, efficiency consumption of the rack PDU itself leads to higher internal
and availability. stress. The amount of heat dissipation within the rack PDU
is determined by its idle power consumption. If the idle
High Temperature Rating power consumption level is lowered, the heat dissipation
Given their location at the back of the rack toward the within the rack PDU is lowered. Idle power consumption
hot-aisle, rack PDUs are exposed to some of the harshest should be an important consideration especially for
temperatures found in the data center. (Figure 2) In this switched rack PDU designs, which provide the ability to
location, it is fairly common to see temperatures of at turn on, turn off or recycle power to connected IT
least 50 degrees Celsius. As power densities continue to equipment through the use of relays at every outlet. Idle
increase and more organizations consider increasing the power consumption is generally not a specification put out
temperature in the data center to reduce energy by vendors and hence should be requested.
consumption costs, these temperatures are only
expected to rise. It’s important that intelligent rack PDUs
are rated to withstand a temperature of 55 degrees
Celsius or higher.

3
 Below are key features that help keep the idle power This means that for a 20A OCP rated at 80 percent, the
consumption of intelligent rack PDUs low. maximum continuous current that it can be used for is
only 16A. An OCP rated at 100 percent would provide a
yy A
 luminum construction - Aluminum has better
conductivity than steel and helps keep the internal maximum continuous current of 20A, making it an ideal
temperature rise to a minimum. It also makes the rack choice to minimize the chances of tripped branch circuit
PDU lighter in weight, resulting in easier deployments breakers due to minor overloads. As always, any circuit
breakers you choose should carry the appropriate agency
yy B
 acklit LCD display - After a time the Display will approvals, such as UL489 in all regions.
automatically switch to a special power safe modus
Smart Inrush Current Management
yy B
 istable relays - These relays, also known as latching
relays, draw power only when there is a change of state Inrush currents are caused by bulk capacitors charging in
outside of normal operation. This helps keep overall power server power supplies, and can be in excess of 50A for a
consumption of the rack PDU significantly lower. A few tens of microseconds. To ensure that any upstream
switched rack PDU offering with bistable relays can save a breakers do not trip, switched rack power distribution units
typical 100 rack datacenter customer up to an estimated
should be considered since they allow the capability to
7600 EUR annually in energy consumption costs.
power up the outlets in a sequential fashion.

Appropriate Overcurrent Protection These high inrush currents can also be detrimental to the
relays within the switched rack PDUs themselves. Key to
For safety reasons, regulatory agencies require rack PDUs smart inrush current management within switched rack
to have overcurrent protection (OCP) above 16 Amps. If power distribution units is to ensure that the opening and
the appropriate OCP is not used in the data center, it closing of the relays is synchronized to be near zero
could lead to incidents of tripping that compromise the crossing of the current / voltage waveforms.
availability of all loads connected to a branch circuit. An
appropriate OCP for rack PDUs should not be highly Input Cable Size
sensitive and should have a minimal mean time to repair
(MTTR). There are different types of OCP devices that can When choosing Wye-connected rack power distribution
be used with intelligent rack PDUs. These include fuses, units, ensure that the input power cord is properly sized to
thermal magnetic circuit breakers and hydraulic magnetic handle neutral currents in the event of unbalanced loads.
circuit breakers.
Power Cord and Outlet Locking
Due to the downtime involved in replacement of fuses, some
rack PDU manufacturers recommend that they not be used Power cord and outlet locking mechanisms secure the
for mission critical facilities such as data centers. If a fuse is physical connection and ensure the power cords are not
blown, it must be replaced – which can be a very time accidently pulled out of the outlet, causing an inadvertent
consuming and expensive fix. In most cases, this involves load drop. (Figure 3) Globally, the most common standard
the upstream breakers on the floor PDU being turned off, a for outlets used in rack PDUs is IEC320 C13 and C19. IEC
step that typically requires a qualified electrician. The receptacles are internationally acceptable and handle
result is significant downtime and a longer MTTR. output voltages up to 250V.
Circuit breakers are better suited for higher density,
higher power consumption applications, primarily because
they can be quickly and easily reset. Thermal magnetic
breakers are designed to trip instantly as soon as the
current threshold has been reached. They are also more
sensitive to ambient temperatures, which does create an
issue given the location of the rack PDUs. Hydraulic
magnetic breakers are more tolerant of current surges
and less sensitive to ambient temperature changes,
making them an ideal choice for intelligent rack PDUs.
Another important point to consider about OCP is the
branch rating. Branch OCP devices found in most rack
PDUs are rated for either 80 percent or 100 percent of Figure 3. Locking outlets and locking power cords prevent
their load. accidental unplugging of IT devices.

4
Functionality Metering
Current sensing within electrical circuits can be provided
Intelligent rack PDUs should be able to provide proactive
through the use of shunts, current sensors or Hall Effect
notification of impending issues before they occur.
sensors. Because shunts sit in the path of high voltage
Warning and critical threshold settings for the current
power, an issue with the shunt itself often leads to
ensure that the rack PDUs do not experience overload
disruption of power within the primary circuit. On the other
conditions that could otherwise trip the breaker and the
hand, current transformers and Hall Effect sensors are
connected loads. While setting the current configuration,
coils that are isolated from the primary high voltage circuit.
care should be taken that in a typical 2N scenario at the
Therefore, a disruption of power to these sensors
rack level, the thresholds for the branches are set at less
themselves has minimal impact on the power flowing
than 50 percent of the overall rack PDU rating.
through the primary path. Current transformers have a
further advantage over Hall Effect sensors in that they
Software Electronic OCP
provide higher accuracy.
Paired with proactive monitoring, this feature will turn off
and lock down all unused outlets on a branch circuit that Switching
has exceeded current established thresholds. It basically
Switching within rack PDUs is vital to allow remote control
prevents someone from plugging new equipment into an
the connected equipment, and is enabled by the use of
unused outlet and causing a circuit overload.
relays at each outlet. The relays used within rack PDUs
Additional parameters that an intelligent rack PDU needs can be one of three types: normally-open; normally-closed
to monitor to ensure high availability: or bistable relays
1. Phase currents, along with a notification of unbalanced yy Normally-open relays require power to be supplied to
loads them in order for the outlets to be able to deliver
power to the connected loads. In the event there is an
2. Temperature within the rack, through integrated issue with the power supply feeding the relays, the
sensors, along with the ability to configure auto turn- relays remain open and there is no power provided to
off outlets when temperatures exceed critical the connected outlets
thresholds
yy N
 ormally-closed relays only require power to open the
3.  Ability to monitor circuit breaker status. (This is outlets. Under normal operation, they remain closed.
typically found in rack PDUs with metering or This means if there are any issues with the power supply
switching capabilities down to the outlet level. For rack feeding the relays, the outlets continue to provide basic
PDUs with metering capabilities only at the branch power distribution to the connected loads
circuit level, low critical threshold could be monitored
as a proxy for circuit breaker status.). yy Bistable relays (or latching relays) act as normally-
closed relays during normal operation, in that they also
All notifications should be capable of being received in a resort to providing basic power distribution in the
standard format, such as SMS, SNMP traps or e-mail. event of a power supply failure. They do provide an
Power distribution units should be capable of integration additional advantage in that they allow a choice of
with a centralized management software, which will enable state when power is restored after an outage. Outlets
them to be easily managed. can either be turned on or off, or returned to the same
state they were in prior to the outage. Bistable relays
Fault Tolerance only require power if they change a state. They will
keep outlets up and running with no power. In addition,
Intelligent rack PDUs should be designed such that a loss their power consumption during normal operation is
of a single phase will not lead to dropped power on all the low, which helps minimize the overall energy footprint
unaffected phases. Also, regardless of the advanced of Switched rack PDUs. (Figure 4)
features offered by an intelligent rack PDU, it needs to
continue to provide basic power distribution in the event External Connectivity
of a compromise of the intelligence capabilities. In the event that the primary network to the rack PDU
Fault tolerance due to a loss of the one of the main goes down, some rack PDUs provide redundant
communications through integration with out of band
intelligence capabilities (i.e. switching, metering and
management devices, such as serial consoles or KVM
external connectivity) is based on the design of these switches.
capabilities.

5
 However, if the external communication to the rack PDU is While the survey provides good data for a more broad
just not available, the design of the rack PDU should ensure discussion on data center downtime and the steps that
that basic power distribution, as well as the operation of local can be taken to increase availability, it does underline the
management modes such as onboard display, is not affected. importance of minimizing the Mean Time To Repair
It is important that rack PDUs have an automated (MTTR) if an intelligent rack PDU goes down.
management path that will maintain basic power distribution.
There are three factors that need to be considered for
An automated management path also will ensure that when
better maintainability.
input power to one of the phases of a three-phase intelligent
rack PDU is lost, the outlets connected to the unaffected 1. Overcurrent Protection Type: As mentioned earlier in
phases continue to be powered. this white paper, circuit breakers are usually resettable
when they trip; while fuses usually require replacement.
The procedure for replacing fuses typically requires an
Normally-open relay: Normally-closed relay:
electrician and requires the input circuit to be turned
off while it is being replaced. Not only does this take
time, it requires coordination with facilities teams

2. Modularity: Modularity along with hot swappability

ensures a faster MTTR. Look for a design with a
modular communications card which can be replaced
30 86 85 87 30 86 85 87
while the unit remains operational and providing basic
Bistable (or Latching) relay: power distribution. Some rack PDUs also provide
modularity of input and output power

3. Switching: Switching capability within a rack PDU

ensures that if connected IT gear hangs up, power can
be remotely turned on, turned off, or recycled without
any physical intervention at the data center. In order to
30 86 85 87 87a ensure that power to the appropriate equipment is
being recycled, care should be taken while associating
rack PDU outlets to IT gear. Rack PDUs that provide
Figure 4. Bistable (or latching) relays resort to providing basic power integration with data center IT equipment access and
distribution in the event of a power supply failure control solutions simplify such association and
minimize the opportunity for errors.
Maintainability Adaptability
As computing demands and complexity in the data center
As a result of the rapid pace of change that is occurring in
continue to rise, unplanned data center outages remain a
the data center environment, one of the biggest
significant threat to organizations in terms of business
challenges today’s data centers face is meeting current
disruption, lost revenue and damaged reputation. A 2013
requirements while ensuring the ability to adapt to future
survey of U.S.-based data center professionals by the
demands. In the past, this was accomplished by oversizing
Ponemon Institute and sponsored by Vertiv™ showed that
infrastructure systems and letting the data center grow
an overwhelming majority of respondents had experienced
into its infrastructure over time. Many data centers are
an unplanned data center outage in the past 24 months
abandoning the oversizing approach because it is
(91 percent).
inefficient in terms of both capital and energy costs.
Regarding the frequency of outages, respondents
This also holds true at the rack level. Rack-level flexibility
experienced an average of two complete data center outages
is an important factor in helping data centers adapt to
during the past two years. Partial outages, or those limited to
constant change, which often means higher densities and
certain racks, occurred six times in the same timeframe.
the demand for greater efficiency and control.
According to survey responses, complete outages lasted an
The need for change can be caused by consolidation,
average of 107 minutes and partial outages lasted an average
moving from one server or network to another server or
of 152 minutes. (Figure 5) The second part of the study
network, or the addition of new equipment.
quantified the cost of an unplanned data center outage at
slightly more than EUR 5800 per minute.
6
 12 11.29 Extrapolated value (incidents)
10.16 for FY 2013
10
Extrapolated value (incidents)
for FY 2010
8 6.84
5.88
6

4
2.04 2.48
2

0
Total DC shutdown Localized shutdown Limited Outages

Figure 5. Frequency of data center outages over two years, based on results of 2013 survey from the Ponemon Institute.

For instance, to handle higher rack densities, there is a Conclusion

need to seamlessly make the change while protecting the
existing investment. Intelligent, adaptive rack PDUs with With increasing data center densities, a single rack can
separate power entry modules provide the ability to now support the same computing capacity that used to
quickly react to this change. In addition, hot-swappable require an entire room. Visibility into conditions in the rack
modular output power ensures that you protect initial can help prevent many of the most common threats to
investment and minimize downtime when the server rack-based equipment, including interruptions to the
architecture within the rack changes. A modular Power power source, accidental or malicious tampering, and the
Distribution Unit (Knürr PowerTrans2), which provides the presence of water, smoke and excess humidity or
interface between the low-voltage power supply and rack temperature.
PDUs, can also be used to support change by providing
Today’s intelligent rack PDUs can provide the best view of
the flexibility to add or modify rack layouts and change
IT power consumption inside the rack. They also can
receptacle requirements without risking power system
provide a level of monitoring and control capabilities that
down time. (Figure 6)
were unheard of just a few years ago. However, to ensure
the technology provides the benefits of a high availability
solution, it’s important that considerations are given to the
design of the rack PDU, the capabilities it offers, its ability
Vertiv™ MPX™ adaptable
rack PDUs to provide faster MTTR, and the level of flexibility it
Liebert APM™ UPS system provides.
Vertiv Knürr Power
Trans2™ power
distribution unit

Feed B
Feed A

Figure 6. Vertiv™ Knürr PowerTrans2™, seen here in combination with

Liebert® APM™ UPS and Vertiv™ MPX™ Rack PDU, can be used to
support power distribution to the rack.

7
VertivCo.com | Emerson Network Power Limited, George Curl Way, Southampton, SO18 2RY, VAT Number: GB188146827
© 2016 Vertiv Co. All rights reserved. Vertiv™, the Vertiv logo are trademarks or registered trademarks of Vertiv Co. All other names and logos referred to are trade names, trademarks or registered trademarks of their respective owners. While every
precaution has been taken to ensure accuracy and completeness herein, Vertiv Co. assumes no responsibility, and disclaims all liability, for damages resulting from use of this information or for any errors or omissions. Specifications are subject to change
without notice.

MKA4LOUKWP