Datacenter-Link Blog

Data center: towards the Information Power Station (First part…)

Alessandro De Danieli — Mon, 08 Feb 2010 08:22:45 +0000

In 1882 Edison switched on the world’s first large-scale electrical supply network that provided 110 volts direct current to fifty-nine customers in Manhattan; in 1883 in Milan, between Santa Radegonda and Agnello streets, started the production the 1st power station among Continental Europe.
The birth of the first power stations was a crucial junction in the evolution of modern society: not for nothing that this event was one of the most important of those that characterized the so-called Second Industrial Revolution.
In his book “The Big Switch: Rewiring the World, from Edison to Google,” (W. Norton & Company, 2009), American writer Nicholas Carr draws a suggestive analogy between the rise of the very large data centers as the ones built recently and the Second Industrial Revolution. Just as nascent industries, once powered by water wheels, began able to run their machines thanks to constant and reliable voltage generated in distant power plants, advances in technology and transmission speeds are permitting computing to function like one utility, a distant but reliable source of services. Really, this is exactly what the CEO of Google, Eric Schmidt meant when last May in one of his speech said to the press that “The Browser is the Computer”.
The metaphor is interesting: we can expand it. Till the 90s, computers used to be stand-alone devices. If you wanted to do something more than looking at the prompt line, you had to buy a software and install it on your PC. Then, the World Wide Web arrived in the late 90s. Suddenly, if you had a network connection and a browser, you could read pages and pages of information not contained in you hard disk. Think about YouTube, MySpace, Facebook, Wikipedia, Google Search, Yahoo Mail: none of those programs is running on one PC’s hard disk. They’re all utility services that everyone can share with people living in every part of the world. And what is very interesting is that no one really cares where our software is coming from, what are the features of this software, in which way it is using the PC: if it works it’s OK. The PC began to be fed by outside, and the focus started to be on the connection.
The analogy is clear: the development of the electric grid move the focus from the need of appropriate source of power to the need of connection to the grid. No one (except for few electrical engineers) is interested into frequency, voltage, power quality… We trust the Net.

…..TO BE CONTINUED…..

Quotes:
[1] http://www.nicholasgcarr.com/bigswitch/interview.shtml, N. G. Carr
[2] “A Practical Guide to the Early Days of Data Center Containers”, October 29th, M. Manos.
[3] “Data Center Overload”, Published June 8th 2009, The New York Times, T. Vanderbilt
[4] “Estimating total power consumption by servers in the U.S. and the World”, February 15th 2007, J. Koomey.

Efficient and Reliable Cooling certified for High Density Servers

Ezio Sturaro — Thu, 28 Jan 2010 16:25:36 +0000

German certification body TUV, has certified that the Uniflair High Density Cooling Solutions are suitable to effectively cool loads of up to 40 kW each rack granting complete reliability and high efficiency of the complete system.The certification process involves the Test Room present in the Uniflair Headquarters dedicated to High density Cooling Solutions having verified the chain of measurement of the equipment installed, acknowledging its precision and reliability.
The ambient was cooled by a Leonardo Chilled Water Precision Air Conditioning unit integrated with AFPS system and the closed racks were cooled by means of Active Floor modules.The Active Floor modules guarantee that the temperature of the air under the floor is lifted to the superior height of the units and to the inside of the rack. Combination with the compartmentalization of the cold corridor is not strictly necessary, even when the system is operating at total capacity.The under floor air discharge temperature can be altered depending on the client’s requirements. According to the project conditions of the chiller, the discharge/return temperature can be designed at an optimum of, for example 12/18, so allowing the use of Free-Cooling for a greater number of hours.The increase by one degree of the discharge temperature from the CRACs allows the use of free-cooling for 300-350 hours per year.Loads of 12, 15, 20, 25, 30, 35 and 40 kW were generated in the test rack by means of resistors. All the loads were safely cooled with a discharge temperature of 20°C. Given that the Active Floor Module works based on the existing installation, the modules which make it up can be used in a rapid and flexible way to resolve the air conditioning problems or to increase energy efficiency, constituting a convenient solution which avoids involving water for cooling of the rack.If compared to “in rack“ cooling this solution, according to the concept, allows a saving in terms of space, guaranteeing a secure investment for datacenter infrastructures cooled through the modular access floor.The first step in existing systems is that of improving the airtightness of the modular access floor and optimization of the under floor pressure.It stands to reason that the cooling of high loads assumes the availability of a corresponding air or cooling capacity in the ambient or rather under the floor.During the tests it was possible to:

effectively cool all the simulated loads (12, 15, 20, 25, 30,35 and 40 kW)
transfer the under floor temperature to the upper part of the unit (with the possibility of increasing the discharge temperature by various degrees).
effectively cool even higher loads with a discharge temperature of 20°C.
select the optimal temperature for the servers in terms of operating safety and/or energy efficiency.•
have available, in the event of an emergency, a more than adequate reaction to eliminate possible breakdowns even in the presence of loads of 40kW.
make the data center’s air conditioning operate in an optimal way
simulate redundancy concepts for high capability data centers to eliminate, in a simple and direct way, the phenomenon of hot spots.

The test examiner expresses a recommendation for the certification according to the catalogue of requisites “TUV certified energy efficiency – High Density Cooling”.

Power Grid Availability

Matteo Granziero — Thu, 14 Jan 2010 17:28:08 +0000

In order for power to be usable (e.g. high-quality), its characteristics must be within the tolerances of the supplied load. Therefore it is not enough for power to be present. Let’s take this necessary condition as the starting point for the analysis.

Power distribution

Information on electrical power quality is somewhat scarse. The main independent sources of information considered here are:

European Network of Transmission System Operators for Electricity (ENTSO-E);
The Italian Regulatory Authority for Electricity and Gas;
Leonardo Energy (LE).

The available statistics respectively refer to High Voltage lines, Low Voltage lines and power quality within the system used. In fact, high-quality power supplies are frequently compromised by consumers’ loads within their own plants, such as non-linear loads or loads with high inrush current. Unfortunately, the lack of information on Mdium Voltage lines, which is of extreme relevance to all consumer loads connected via transformer to the grid, does not allow for a complete analysis. The reason probably stems from the fact that individual quality standards are defined in Medium Voltage.
The ENTSO-E Statistical Yearbook 2008 provides an overview of electricity exchanges between European nations and lists the major periods of unavailability, classified according to the reason:

Maintenance
New construction
Overload
Failure in protection devices
Outside impacts (animals, plants)
Atmospheric agents
Other reasons/unknown

The bottom line is that in 2008, the total downtime of international tie lines of 220 to 380 kV with rated power from 300 to 1745 MVA between Italy and neighbouring countries amounted to approximately 6500 hours.
However as regards low-voltage lines, the Regulatory Authority informs us that in 2007 in Italy, each user experienced 4.73 power outages of between 1 second and 3 minutes and 2.16 power outages of over 3 minutes’ duration, giving a total downtime of 58 minutes.

Online the Leonardo ENERGY Data Centres Micro Site

Matteo Granziero — Fri, 13 Feb 2009 09:55:59 +0000

Leonardo ENERGY is the premier web site delivering a range of virtual libraries relating to electrical energy.

It is an initiative dedicated to build information centres to serve designers, engineers, contractors, architects, general managers, teachers and students, professionally or otherwise involved with electrical power.

Leonardo ENERGY has developed a micro site dedicated to Data Centres complementary to Datacenter-Link.

Protection and switching of transformers

Roberto Pomari — Mon, 20 Oct 2008 13:08:45 +0000

Transformers are used to achieve a change in the supply voltage, for both medium and low voltage supplies. The choice of the protection devices must take into account transient insertion phenomena, during which the current may reach values higher than the rated full load current; the phenomenon decays in a few seconds.

The curve which represents these transient phenomena in the time-current diagram, termed “inrush current I0”, depends on the size of the transformer and can be evaluated with the following formula (the short-circuit power of the network is assumed to be equal to infinity)

where:

K ratio between the maximum peak inrush current value ( I0 ) and the rated current of the transformer (I1r): (K= I0 / I1r);

τ time constant of the inrush current;

_Ir1 rated current of the primary;

t time.

The table below shows the indicative values for t and K parameters referred to rated power Sr for oil transformers.

Further to the above consideration, the follwing diagram shows the inrush current curve for a 20/0.4kV of 400kVA transformer. This transformer has an inrush current during the very first moments equal to about 8 times the rated current; this transient phenomenon stops after a few tenths of a second.

What does Power Factor mean in a UPS?

Matteo Granziero — Wed, 15 Oct 2008 09:07:55 +0000

If someone looks for Power Factor (p.f.) value in a UPS data-sheet, he will find two values: input and output.

What do they mean? Isn’t the load that defines the p.f.?

Focusing on the IT loads even if the concept is general, the first thing to understand is that the server is the load for the UPS and the UPS is the load for the utility. Therefore the input p.f. represents the the way the UPS affect the utility and allows to chose properly cables, breakers, PDU, etc.
An example can clarify the concept. Let’s compare two three-phase UPSs with input p.f. 0.99 (1.) (IGBT technology) and p.f. 0,68 (2.) (six pulse SCR rectifier) both in the worst condition for the mains, discharged batteries, supplying a 180 kW p.f. 0.9 load. Neglecting for a moment the efficiency it gets:

Pn=180kW, Sn=180kW/0.99=182kVA, I=182kVA/(400Vx√3)= 262A;
Pn=180kW, Sn=180kW/0.68=265kVA, I=265kVA/(400Vx√3)= 382A.

In the second case a p.f. compensator banks is needed increasing the plant cost.
Please, notice the the use of the UPS 1. gives benefits even comparing with the load directly connected to the facility:

Load: Pn=180kW, Sn=180kW/0.9=200kVA, I=200kVA/(400Vx√3)=288A.

Then, what does the UPS output p.f. mean?
It is the the p.f. of the load that the UPS can supply without derating. Consider that both the UPS output limits kVA and kW can not be exceeded and that the output p.f. is given by the ratio kW/kVA.

In this case two examples are necessary.

Example 1

Load 180kW p.f. 0.9 (Sn=200kVA)

Output p.f. 0.9 for UPS 1. and 0.8 for UPS 2.

Pn=180kW, Sn=180kW/0.9=200kVA
Pn=180kW, Sn=180kW/0.8=225kVA

In case 2. a bigger UPS is requested to supply with consequent cost increase.

Example 2

Load 180kW p.f. 0.8 (Sn=225kVA)

Output p.f. 0.9 for UPS 1. and 0.8 for UPS 2.

Pn=180kW, Sn=180kW/0.9=200kVA
Pn=180kW, Sn=180kW/0.8=225kVA

In case 1. the UPS is overload and after few minutes it would switch to bypass load supply. This happens because the UPS Sn limit is exceeded.

Therefore there is no the perfect output p.f. but the proper one.

It is important to say that in the IT application the servers’ p.f. is typically 0.9 leading.

Conclusion: choose the UPS with the highest input p.f. and the proper output p.f.

Protection and switching of generators

Roberto Pomari — Tue, 02 Sep 2008 10:31:40 +0000

The need to guarantee an ever greater continuity of service has led to an increase in the use of emergency supply generators, either as an alternative to, or in parallel with the public utility supply network. Typical configurations include:

“Island supply” (independent functioning) of the priority loads in the case of a lack of energy supply through the public network;
supply to the user installation in parallel with the public supply network. Unlike the public supply network, which has a constant contribution, in case of a short-circuit, the current supplied by the generator is a function of the parameters of the machine itself, and decreases with time; it is possible to identify the following successive phases:1. a subtransient phase: with a brief duration (10÷50 ms), characterized by the subtransient reactance X”d (5÷20% of the rated impedance value), and by the subtransient time constant T”d (5÷30 ms);2. a transitory phase: may last up to some seconds (0.5÷2.5 s), and is characterized by the transitory reactance X’d (15÷40% of the rated impedance value), and by the transitory time constant T’d (0.03÷2.5 s);3. a synchronous phase: may persist until the tripping of external protection, and is characterized by the synchronous reactance Xd (80÷300% of therated impedance value).

Virtualization applications and physical machines shut down

Damiano Buscemi — Fri, 25 Jul 2008 10:03:15 +0000

Server virtualization in Datacenters is one of the most implemented solution to make optimized utilization, improve efficiency and reduce consumption of the phisical infrastructure. While it is clear and well known to everybody what are the advantages of Virtualization from the IT point of view, the effects of virtualization on the physical layer of the datacenter are less popular. While having one single physical machine running one server means having one single point of failure (the machine) for each server, from an availability point of view, having multiple independent operative system (servers) running in one single physical machine, makes a larger number of servers dependend on one single point of failure (the machine). There are more Virtual servers than physical servers.
From this point of view it is more and more important to make the physical machine able to perform the necessary operations, for instance the shut down, while the multiple Virtual servers are running on it, and when critical power quality conditions exist.
UPS manufacturers for instance provide different client softwares which receive signals from their machines when the critical power event happens. Most of these Shut down clients are designed for traditional servers shut down. At this point we have to remember that between the phisical machine and the virtual servers there is an hosted or non Hosted application which needs to shut down after al the Virtual machines are turned off. It is more and more important to have clients able to shut down these applications which virtualize the hardware for each Virtual server, for example the VMware ESX or the Microsoft Virtual Server 2005 and perform a safe and complete software and hardware shut down.

Automatic Floor Pressurization System

Ezio Sturaro — Fri, 11 Jul 2008 10:57:48 +0000

Most technological environments use an under floor distribution technique to maintain the conditions within the built environment.
The principle is simple and long established and uses the pressure underneath a raised access floor in order to ensure that cool air is available wherever an air outlet (usually a grill) is positioned. Maintaining good pressurisation is important in order for the air conditioning system to work efficiently. This aspect must be guaranteed for the entire life span of the room and be able to be modified over time.
The AFPS system (Automatic Floor Pressurization System) developed and tested by Uniflair ensures automatic adjustment to the air flow according to the servers installed enabling flexible installation regarding the infrastructure.
The AFPS ensures automatic adjustment of the air flow issued by the perimeter units with EC fans during ordinary andextra maintenance to maintain a constant under floor pressure by maintaining precise control of the air distribution / cooling in all of the room (eliminating Hot Spots). In fact, during ordinary maintenance, access raised floor panels are often replaced which therefore reduces the static pressure underneath the floor. As a consequence, the air flow issued by the grills is reduced and the risk of hot spots developing is increased. The control module, which can also be used with electronically commutated fans, allows a nominal pressure to be maintained underneath the access raised floor (from 20 to 80 Pa) and to manage the fan speed ensuring that the nominal pressure value (which can be set) is maintained during all of the operation phases of the unit during the life span of the room itself.
The system is composed of the following main elements:
1. Precision air conditioning units featuring modulating fan control (by means of electronically commutated fans);
2. Microprocessor regulation system with dedicated regulation software;
3. Pressure transducer which can be installed underneath the access raised floor and which is able to monitor the static pressure;
4. Pressure sensor with anti-fouling and “filtering” system of the moving components;
5. An assembly system for the pressure transducer ensuring reliable readings which are not influenced by dynamic effects;
6. A communication and management system of the LAN parameters integrated in the microprocessor control of the perimeter units.
The system manages the variations in pressure underneath the floor by means of an integrated system of automatic pressure regulation in order to deal with any eventual changes which are too rapid, therefore stabilizing the system. The system also manages the constant pressure underneath the floor during ordinary and extra-ordinary maintenance of the floor and also when new servers are installed, adapting the flow when:
• New equipment is added;
• The floor panels are opened during maintenance / installation of new equipment (without creating hot spots in another point of the same room);
• Partition walls underneath the floor break or are damaged.
The system can be integrated both with chilled water and direct expansion air conditioning units. In direct expansion units, the management software must allow for dedicated air flow regulation settings. The system is able to manage all of the information read by the different units and define combined regulation strategies by means of a LAN connection (Local Area Network); The system is able to manage the air flow of both a single unit as well as all of the connected units to ensure that the pressure underneath the floor remains constant; The system is able to define the nominal pressure value which is requested via the setting of the microprocessor control. The system is able to read the average pressure value of a specific area (one for each unit) with its management logic. The system may have a single point of reference in the room or may be managed according to various areas. In the second case, it is possible to control all of the units based on the average pressure read by all of the units with the exception of the areas in which the pressure “differs” too much from the average value. In this case, the units within this area manage the air flow independently in such a way that this specific single area also returns to a nominal value. The system manages the growth of the room over time: automatically changing the cooling capacity and the air flow depending on the number of units, grills and air distribution systems added.
The possibility of managing the air flow according to the growth of the room enables the absorption due to the fans to be reduced; in fact, when the room is not complete, the air flow needed is less than the nominal value, the AFPS system partializes the EC fans with significant benefits in terms of absorption. When there is a unit in stand by, it is recommended that it is kept switched on in order to optimise the energy efficiency , above all at partial loads.

Automatic Transfer Switches

Roberto Pomari — Wed, 18 Jun 2008 10:21:04 +0000

In the electrical plants, where a high reliability is required from the power supply source because the operation cycle cannot be interrupted and the risk of a lack of power supply is unacceptable, an emergency line supply is indispensable to avoid the loss of large quantities of data, damages to working processes, plant stops etc. For these reasons, transfer switch devices are used mainly for:
• power supply of hotels and airports;
• surgical rooms and primary services in hospitals;
• power supply of UPS groups;
• databanks, telecommunication systems, PC rooms;
• power supply of industrial lines for continuous processes.
ATS010 is the solution offered by ABB: it is an automatic transfer switch system with micro-processor based technology which allows switching of the supply from the normal line (N-Line) to the emergency line (E-Line) in case any of the following anomalies occurs on the main network:
• overvoltages and voltage dips;
• lack of one of the phases;
• asymmetries in the phase cycle;
• frequency values out of the setting range.
Then, when the network standard parameters are recovered, the system switches again the power supply to the main network (N-Line). ATS010 is used in systems with two distinct supply lines connected to the same busbar system and functioning independently (“island condition”): the first one is used as normal supply line, the second is used for emergency power supply from a generator system. It is also possible to provide the system with a device to disconnect the non-priority loads when the network is supplied from the E-Line. ATS010 device is interfaced by means of appropriate terminals:
- with the protection circuit-breakers of the N-Line and of the E-Line, motorized and mechanically interlocked, to detect their status and send opening and closing commands according to the set time delays;
- with the control card of the Gen set to control its status and send start and stop commands;
- with any further signals coming from the plant in order to block the switching logic;
- with the N-Line to detect any possible anomaly and with the E-Line to verifythe voltage presence;
- with an additional device to disconnect non-priority loads;
- with an auxiliary power supply at 24 Vdc ± 20% (or 48 Vdc ± 10%). This supply source shall be present also in case of lack of voltage on both lines (N-Line and E-Line).

This scheme shows a plant having a safety auxiliary power supply.