|
|
|
1. Actively
cooled server rack cabinet for the lower power segment
Seidl Technologies is looking for one or more partners, preferably but not necessarily from the European
Union, to launch a product, which could be dealt with under the title Energetically
reasonable one-cabinet data center with extremely low maintenance
requirements. With such a product, we would like to fill a gap. The underlying idea is not fundamentally
new. New is the focus, which is set here. As the analysis of the products of well-known manufacturers shows, there is little agreement in this regard.
So we want to establish an Actively Cooled 19-inch Server Rack Cabinet
in the market that is particularly suitable for small companies, medical offices, law and engineering firms,
perhaps also for studios, and, that has a good price. In the case of smaller companies, the catalog of
requirements with regard to such a product looks basically different from that of larger companies or data
centers. In general, it is seen that, in smaller companies, it is accurately explored that an imagined
procurement will advance some matter of importance and will do that for a very long time.
2. Needs
and solutions from the technical point of view
It is not always right to give data and their processing out of the hands. When it really comes down
to it, it is necessary that data protection is physically enforced. This in its turn requires reliable,
own hardware that cannot be accessed by strangers. The opinion about what one has to understand under
reliable hardware changes with time, more precisely with the technological development. What, on the
other hand, seems to be relatively steady is that reliably appearing hardware comes typically in the
19-inch format. And that is first of all a good thing.
IT equipment in the 19-inch
format can excellently be planned, installed and maintained. It is also undisputed that equipment in this
format saves space, it is firstly stacked on top of one another and only then juxtaposed. Moreover, it is
considered to be comparatively reliable, not necessarily because it works much better, but because it is
usually uncompromisingly ventilated and most of the cables and connectors are protected from direct access.
On the other hand, we have a number of drawbacks. Installed in office environments, 19-inch standard cabinets being full of equipment, more precisely,
standard cabinets with significant energy conversion are perceived as very loud. Besides, they get polluted
with time in a very unpleasant manner. Both the noise and the pollution are results of the strong and
continuous gas exchange with the environment. The market takes neither the noise problem nor the dust one
seriously. This is not really surprising. Well, in the huge halls of the data centers, the noise does not
really matter and there is also no dust there. And since equipment, which is not intended for a data center,
normally does not come to the customer in the 19-inch format, but
in any cases of any size, geometry and color, and usually with a wall plug power supply, actually, there does
not seem to be any reason to think about how a quiet and clean 19-inch
cabinet should be built. This can be seen quite so. Furthermore, equipment encased in 19-inch standard cabinets gets quite fast unstable or dies the heat death
when such a cabinet was placed in a possibly windowless broom chamber and someone has closed the door, even
though the air inside there circulates well. Even a possibly existing air conditioner can only change this
situation if the generated heat is actually removed from the chamber. But this is exactly that what is not
done by products with a compact air conditioning unit. In addition, IT
equipment in the 19-inch format costs significantly more than
functionally equivalent one in a plastic or sheet metal box. And finally, there is interesting equipment
that is not available in the 19-inch format at all. In spite of all
the mentioned disadvantages, in 2005 we built an air-cooled cabinet, so to speak, as a first attempt,
whereat it was initially not thought about its commercialization. This cabinet is still in operation today.
The most important experiences and conclusions are the following ones.
- The massively made is completely closed except for a dust-filtered intake and an outlet opening. The walls consist of 26 mm oak wood, 12 mm cleaned bitumen, 5 mm plywood, and 3 mm neoprene, from outside to inside. Front and rear doors are
screwed in place and sealed with neoprene. Vibrations of the air column in the exhaust duct are damped
by means of an acoustic labyrinth. If the two openings were plugged with, for example, pillows, the sound
insulation would probably be perfect. Unfortunately, the latter is not possible with air cooling, so that,
with the outflowing medium, some of the noise escapes, too. The air filter in the intake duct is ⅓ m² in size so that it need not be changed or cleaned
too often. Suitable filter mats are those of class F9 and of class F7 as the prefilter. At continuous fan
operation, a filter change is required in the countryside every two years, and in a southern European city
with a high traffic volume, one every year. An air filter that looks like has not been serviced for too long, whereat the photographic color contrast has been raised somewhat
for clarity. The cabinet accommodates an uninterruptible power supply (UPS), three 4U rack servers, a 4U special rack chassis being full of small equipment, and a plethora
of cables. Thus, the cabinet firstly fulfills its most important function. On the other hand, it is seen
that we are already at this point in a position to draw conclusions of some relevance. With regard to
commercialization, a cabinet in the form of a piece of furniture will likely be unaffordable and also
hardly approvable. So a 19-inch standard steel cabinet seems to
be the best candidate, which could also quite easily be anti-drummed by
means of panels of cleaned bitumen. Further, air cooling is generally unfavorable. Neither of the two core
problems, noise and dirt, can be dominated comfortably with air cooling. A commercializable layout will better
work with water cooling. Therewith, the IT equipment is still cooled
with air, and water as the medium serves to remove the altogether generated heat from the cabinet. With
water cooling, a massive gas exchange between the cabinet and the environment need no longer happen. So,
sound deadening can cost-effectively be achieved, so, the problem
of getting polluted is basically no longer an issue, and so, the broom chamber situation has its natural
solution. Toward the customer, water cooling is the only honest approach. He buys a product and has the
real chance to be at rest for the period of a decade. This is not conceivable with air cooling in an
office environment
- If such a cabinet is reasonably designed, it will easily reach a life of the order of 25 years. Wearing parts are firstly the fans. For models with
two ball bearings, several manufacturers give 80,000 operating
hours at 40°C. This lifetime will actually be achieved with
high probability, since these fans are, even in the air cooling case, operated virtually dust-free, provided that the above mentioned air filter is appropriately
maintained. So far, not any one of the more than 40 fans has failed.
At a rough estimate, they have been cleaned once every 4 years with
a paintbrush and a vacuum cleaner, whereat the respective contamination was indeed already visible in color
but still harmless. As already indicated, in a cabinet with water cooling the fans need no longer be cleaned.
Customers who depend on that will preventively replace fans that have 80,000
operating hours or 9 years behind them, regardless of whether these
fans belong to a redundant design or not. Finally, in a water-cooled
cabinet, the circulator pump will also have to be counted as a wearing part, whereat it can quite strongly
be benefited from the considerable technical advances of the house technology products in recent years.
A modern pump will therewith last at least a decade
- Unlike in a data center where practically everything runs all the time, the equipment in the
application environment of our target customer group is, due to the high electrical energy prices,
frequently switched on and off, and this on a per-device
basis mainly governed by the day-night cycle. An online UPS
according to the double-conversion principle, which itself
favorably runs continuously, is very suitable for such operating habits. Firstly, the double-conversion UPS products of all well-known manufacturers are considered to be extremely reliable
on their own. As well, on the condition that such a UPS is reasonably dimensioned, an advantage,
which should not be underestimated, is that the components powered by it enjoy the virtually eternal
life. So, each time some device with large capacitors is switched on, the UPS goes completely down
into its knees for self-protection, and, in doing so, it keeps the inrush current of the powered
device within relatively narrow limits. With an online UPS, the probability increases to get
devices, having switched off last evening, back into operation next morning. In our cabinet,
the UPS performs central functions. Consequentially, the choice of the actual type will
hardly be left to the customer.
Eligible models here are 2U or 3U high, have a power of
around 1 kW and work completely autonomously, id est, without a dedicated server. They have apart from the group
of the master sockets at least 2 groups of independently
switchable slaves on the output side. If those slaves are missing, a separate power distribution
unit (PDU) need be used, which makes the entire installation
unnecessarily more expensive. Having active power factor correction
(active PFC) on the input side should not be discussed anymore. Possible UPS models are
equipped with a network card with temperature and humidity sensors as well as with at least 2 dry contact inputs that can be evaluated through software.
So the UPS can, in the case of water cooling, via appropriately connected sensors, directly react to
cooling water circuit disasters, such that it is not necessary to wait for the resulting aftereffects,
namely, for the occurrence of dangerous temperatures. The network card should support status queries
based on well-documented and widespread network protocols
such as the SNMP with the standard UPS MIB according to RFC 1628. If this point is ignored, some trouble with the
appropriate software on the servers is inevitable. Because wherefrom should a UPS manufacturer know
which operating systems will run on the powered servers someday. The web server running on the network
card for the administration of the UPS should support TLS-compliant
encryption protocols. Furthermore, interesting UPS models have, in the best case, installed 3 or more AGM lead batteries 12 V 7 Ah of the size 151×94×65 mm³ with Faston Quick Terminals 250 (F2 (6.3 mm)). These batteries can economically
be exchanged. Finally, a very crucial condition that a UPS has to meet here is that it can be turned on
and off via a dry contact. As a reminder, the UPS is located in the cabinet and is thus only accessible
to a limited extent. Meanwhile, there are at least two different candidates on the market, which fulfill
the whole spectrum of requirements here, id est, UPS models that
one can build on
- There are different ways to deal with several servers. In the air-cooled cabinet shown above, there are three systems, a master server,
a master server cold standby, and an installation server. The master server is the workhorse, the master
server cold standby is a replica of the workhorse in cold standby mode, and the installation server is a
machine on which operating systems are prepared and which is always taken when a sandbox is needed in hardware.
In the case of such an arrangement, two of the three systems are usually switched off. Further, there are
various ways to access central resources, and such ones are the systems in our cabinet right now. If there
is a demand for high safety standards or the need for limited expenditures in administration, employees
access a server from their workplace via a thin client or a diskless workstation. With full confidence, a
person gets the console cables installed directly to the desk. This person then has the full performance
without having to accept the disadvantages which are usually associated with large machines in the immediate
vicinity of the workplace. In the above case, the consoles of all servers are concentrated using a KVM switch
and connected to a writing desk using extender equipment to have the 15 m long cables under control
- For the servers in the above air-cooled cabinet, three 19-inch rack mount enclosures of the size 4U × 700 mm were originally selected, each of which
has three 120 mm fans at the internal partition panel. This type of
enclosure corresponds to a full tower case lying on the side. It is therefore capable of accommodating
all possible standard power supplies and mainboard formats as M-ATX, ATX,
E-ATX, SSI CEB and SSI EEB, but in contrast to a full tower case, it accomplishes
uncompromising ventilating conditions. The hard disks are located in the cold air area behind the two
front doors with their additional air filters. The three 120 mm
fans at the internal partition panel, with their key parameters 60 Pa at dV/dt=0 and 3×50 ℓ/s at Δp=0, care for pressure
differences of at least 40 Pa under all circumstances of
interest here, so that the takeouts at the rear panel run quasi at Δp=0. Under such conditions, a takeout of the dimension 80×80×20 mm³ at the rear panel with 7 blades
of 45° slope and 3300 rpm conveys, just from its geometry, anything like 25 ℓ/s or 90 m³/h. The server enclosures in the above air-cooled cabinet have only one takeout. Two would reduce the air
temperature rise to half. However, the chosen enclosure type with an electrical expenditure of
significantly more than 20 W per chassis just to supply
the fans is even without the second takeout hard on the limit of the tolerable. In case of a water-cooled cabinet, some things will also become more easy in this
context. Resulting rack mount enclosures are first of all similar to the chassis used for servers in
data centers.
In the further process, the three 4U servers in the
above-mentioned air-cooled
cabinet were replaced in 2018 after 15 years of operation by two industrially manufactured used
2U servers. This enabled the available system performance to be
multiplied in every dimension at almost constant heat dissipation. Hence there will be nothing to do for a long time on the
hardware side. Each of the two servers was purchased for a good € 300
and upgraded with further € 1000. As a result, there are two machines
available, each with 12 cores and support for 24 threads at a maximum of 2.67 GHz,
48 GB RAM and a logical RAID 6
hard drive of 2.4 TB.
Two of the eight physical hard drives of each machine are hot standbys.
In the meantime, the passively cooled graphics card for 2560x1440 at 60 Hz
has been replaced by a Radeon Pro WX 5100 for 3840x2160 at 60 Hz, with the 43-inch 4K monitor connected
via a hybrid DisplayPort cable made of fiber optics and copper. Unfortunately, this means that the video signals can no longer be
pulled through a KVM switch. In addition, the eight quite old physical SAS hard drives of the master server, with a capacity of
600 GB each, were replaced by eight used
PM1635a 960 GB solid-state drives for a total price of
€ 800, so that the
I/O performance has now grown to an appropriate level.
There is USB 3.0 for hard drive backup, also via a hybrid cable.
Of course, as expected, well-programmed USB 2.0 devices have no problems with such hybrid USB 3.0 cables, too. USB 3.0 is also good for external sound cards.
Some USB 2.0 sound cards do not work properly on USB 2.0
because they try to draw too much electrical current.
The servers have been designed by the manufacturer for warm data centers. They have regulated fans and can be operated at air
temperatures in the range from 10°C to 35°C.
The presence of a console processor is also advantageous. As usual, installing the latest firmware requires some experience,
but since the systems represent widespread, well-documented hardware of a
famous manufacturer, one does not face serious problems. The same applies in principle to the procurement of spare parts.
Figure
shows below a system as it was delivered at some point in time for a good € 300
and above that a system after its upgrade. One can see that above the optical drive is missing and the hard drives moved to the
drive bay in the middle. In the left drive bay, the hard drives on the left became impermissibly hot, even when the enclosure
was operated in the free space outside the cabinet. Why the latter is the case becomes clear relatively quickly by taking a
from the top.
One last figure, , completes the overview of the servers.
It should be mentioned that Reverse Flow Blockers can practically only be mounted on the front side when using industrial servers
- The mentioned special rack chassis allows to accommodate small non-19-inch devices such as routers, switches, media converters,
KVM hardware, private automatic branch exchanges, parts of video surveillance devices, et cetera, together with their many wall plug power supplies, and,
it allows to operate them under extremely favorable conditions. Even Wi-Fi access points can be hidden away in this manner, if only their
antenna cables are routed out of the cabinet again. In this context, the experiences speak a clear
language. Small devices, which have found their place in the special rack chassis, do not show the
slightest instability. For that matter, the actual ambient temperature is largely insignificant. The
point is the permanent ventilation such that temperature hotspots are avoided. Without such a special
rack chassis, the envisaged target customer group will hardly be enthusiastic about a 19-inch cabinet. Only if a cabinet succeeds to absorb virtually the
entire portfolio of available IT equipment in an expedient and discreet manner, then it fulfills its
purpose here. As already denoted, expedient means in this case not only to house IT equipment, but also to ventilate it well and dust-free. The special rack chassis needs to be dextrously designed.
It was found that the height of 4U is not enough. On the other hand, a height of 8U appears optimally.
This chassis must be accessible from the front and from the rear without having to be pulled out. It
must be sensibly divided, must offer numerous possibilities to fix the small devices and their cables,
and must provide many mains sockets at a sufficient distance from one another. It makes little sense
to mount the special rack chassis movably in the cabinet. The large number of cables, which finally
enter or exit at the rear side, is likely to prevent at the end any movement of the chassis. Lastly,
just to complete the understanding. Hardware that requires operator interventions, as for example
printers, can neither have a place in the special rack chassis nor in the cabinet at all. Similarly,
such things as ADSL modems should also not be dragged into the cabinet. It is safer to have a purely
optical connection between a far away modem and the cabinet, in order to keep the overvoltage problem
under control
- The height of 18U provided by the air-cooled cabinet
shown above is too small. So it is quite unpleasant, if, in addition to some job to put a new machine
into the cabinet, respecting time requirements, and to set it into operation, the decision has to be made,
which of the previously installed servers has to be forever dispensed with. It is much more comfortable
to have both things decoupled from each other. In this context, 42U, for example, offer a luxury of
quite another kind. In addition to the customarily used hardware, there is a lot of space that could
serve, on the one hand, to hold retired servers as standby systems, data storage or sandboxes available,
and, on the other hand, to store unused units in a space-saving
manner, and that at least as long until it is clarified that they are actually no longer needed
- As already indicated, it is assumed that, due to the high electrical energy prices,
the target customer group being in our focus switches the equipment on and off on a per-device basis, mainly governed by the day-night cycle. This creates another problem. Thus, if
no action is taken, units that are switched on can overheat because the cooling air can
partially flow back through units being switched off such that a running unit does not
exclusively get cold air at the inlet, but a mixture of cold air and its own exhaust air.
Depending on the instantaneous flow conditions, dangerous temperature increases can occur
that way with disastrous consequences. Although the air-cooled
variant shown above may be more susceptible to this effect than a water-cooled variant, it is in both cases necessary to
provide a remedy. The technological solution here is to install the already mentioned Reverse
Flow Blockers. In data center environments, this issue plays a subordinate role
Let us take into account all points touched. Our new cabinet, whose commercialization is intended,
will be designed to provide a safe and unobtrusive home for almost everything of remotely controllable
equipment being available in the IT market, to allow the housed
equipment operating under favorable conditions, to run, apart from due UPS battery changes every three
to five years, maintenance-free for something like a decade, to
represent an energetically reasonable solution, to have a generally accepted price, and to enjoy itself
a very long life. A complete package consisting of four apparently separable components wants to be
delivered.
Three of these four components are schematically shown in the above illustration. On the left
we see the cabinet. It could be a symmetrical 19-inch
standard cabinet, made from steel, with closed doors in the front and rear and the dimensions 42U × 600×1200 mm². The
spacing between the front and rear mounting posts should be 800 mm. IP 54
is planned as the degree of protection. An optional anti-drumming kit should be offered to improve sound absorption.
In the cabinet, the 8U air-water heat exchanger is located
below. Above it there is the pump casing. On the right side of the illustration we have the water
recooler. In our case, this is not a recooler with a chiller, that is a recooler with a machine
performing thermodynamic cycles, but a common warm water convector radiator for living quarters,
indeed a model of relatively high power. The third visible component is the tubing. With ½-inch armored hose, there is a pressure drop of
approximately 0.3 bar at 5 ℓ/min volume flow rate and 40 m length. In approximation, the pressure drop increases
with the square of the volume flow rate and linearly with the length. For further orientation, the
water temperature rise is little less than 3 K at
5 ℓ/min volume flow rate and 1 kW total dissipation power. The fourth and last
component, a small control device, is not represented in the illustration.
It can be seen what should happen here. The problem case, the standard server rack cabinet with
its characteristics large, loud, heat and dust sensitive is smashed into two parts, which can more easily be
controlled. A water-cooled server rack cabinet can be placed
practically everywhere, in particular into the mentioned dark broom chamber, whose door could be
specially secured for this purpose. Within the cabinet, comparatively high air volume flow rates can
be realized to care for smallest air temperature rises. Concerning the space consumption, practically
nothing changes. The cabinet will be aligned with one of its side panels towards a wall. Before and
behind the cabinet, the necessary maneuvering space must, now like before, be reserved, which, however,
can be occupied by easily movable objects since the experience shows that a server rack cabinet need
be opened less frequently than once a year. With 21U for servers and storage, there is quite an amount
of space available such that different concepts can be considered. One point is thereby, of course,
that everything can be switched on and off from the outside. The warm water radiator will, on the other
hand, be installed where it is most appropriate. It makes no noise and heats the room which it is in.
It should therefore not be covered.
There will be several questions. A first is surely the one of the reason why a warm water radiator
should be used here instead of a water recooler with a chiller, or, again, instead of a water recooler
with a machine performing thermodynamic cycles. For the intended solution, two circumstances speak first
of all. Firstly, it would be very difficult to communicate to our target customer group that there is a
need for a device, weighing 30 to 70 kg and put-putting permanently or intermittently around, to
properly operate some servers that should replace the up to now somehow put up computers of a customer.
And where, if necessary, should such a device find its place in an office, perhaps in the kitchenette?
Secondly, in a data center with its enormous power densities, there is hardly any other way than to spend
one more amount of electrical energy for cooling that is almost of the same order of magnitude as the one
being used to supply the equipment itself. The recoolers used there are at the same time much more
efficient than any 1 kW model. In addition, data centers
as large consumers are relatively favorably tariffed, whereas the final result of the German energy policy
for the consumer is a kilowatt-hour price for electrical energy around ⅓ €. One consequence is that a usual tower case
computer installed in an office must not consume more electrical energy for its cooling than that for a
few fans. The same rule should basically be applied to the servers in our cabinet, too. Such an argument
speaks even more against a water recooler with a chiller. In addition, the situation can be seen from a
completely different angle. IT equipment transforms electrical
energy into heat. In many parts of the world, this heat can almost all the year round be used to heat
the rooms, which means that the electrical energy is used to turn bits in the processors as well as to move
bytes, and, on the other hand, to heat the house. So the balance is better than is commonly assumed, but
of course only with the warm water radiator. In other parts of the world with perhaps 45°C outside temperature, IT
equipment can normally not be operated without the support of refrigeration machines. But also in
these areas, it can be sensible to get rid of the heat through a warm water radiator directly to the
office, which is often cooled down to 18°C. It is thus
possible to dispose of the heat without great circumstances via the building air-conditioning system. So
there is something that speaks for the warm water radiator with its two great advantages, not making
noise and being almost maintenance free. Nevertheless, we will design the air-water heat exchanger in such a way that condensate drainage is
perfectly mastered, too, so that the cabinet can also be operated with a recooler with a refrigeration
machine. Those customers who want to cool in this manner will, of course, abdicate the pump casing
over the air-water heat exchanger.
Further questions arise. Probably the most important point is to find out under what conditions the
whole installation really works. For that we apply a method called Design of
Experiments with its brute-force variant Full Factorial Design, whereat the experiments are replaced by computer
simulations. Results obtained in this way can be illustrated in a special graph. First of all, however,
is to be clarified what is calculated. The cold air flow indicated on the left in the above illustration
is uniformly distributed over a very large number of imaginary, identical rack mount drawer units of
infinitesimal height. A first, binary influencing factor determines whether the cold air flow is only
distributed over the units being switched on, or over all the units, where a maximum of 10⁄13 of all units can be switched on. The air volume flow inside the
units being switched on is selected via a further, binary influencing factor such that the air temperature
rise equals either 3.3 K or twice that value. The volumetric
flow rate through the units being switched off is, as long as it differs from zero, the same as the one
through the units being switched on. The air, emerging from the units on the right, gets mixed and yields
the warm air. The calculations concerning the air-water heat
exchanger are controlled by means of two further influencing factors. A first, again binary one determines
whether the quotient of the temperature difference between cold air and cold water for cocurrent flow and
the same temperature difference for countercurrent flow should equal 1.82
or 1.42 . The air-water heat exchanger itself is, of course, operated
countercurrently. The second, ternary influencing factor determines the contact surface of
the air-water heat exchanger. The latter can be 2A, 3A, or 4A, where A is
slightly more than 5m². A further influencing factor
determines the volume flow rate of the cooling water, 150, 300, or
600 ℓ/h. Finally, a last, ternary influencing factor determines the nominal
power of the warm water radiator. This can equal 3, 6, or
12 kW. The following graph shows the most important results.
Shown is the very important difference between the temperature of the air, exiting one of the
switched-on rack mount drawer units on the right, and the
ambient temperature at the installation location of the warm water radiator, and this temperature
difference above the overall consumed electrical power being transformed into a heat flow rate out
of the cabinet. Parameters are the mentioned influencing factors. For the maximum permissible value
of the temperature on the output side of a switched-on rack
mount drawer unit, we take the value of 40°C, which was
once given by Intel for the maximum permissible operating temperature of mainboards. This value is
not irrefutable, but it coincides with long lasting experiences with consumer-level products made by different manufacturers. In short,
up to 40°C board temperature, there is not the slightest
complaint in terms of stability when using ECC memory. Further, the offices of Seidl Technologies,
as an example, have south-facing windows and are not air-conditioned. Therefore, the temperature there can just rise
once to 30°C in summer. If one of these offices is now
also the location of the warm water radiator, then all permissible operating conditions of the
installation can directly be read out. Thus, at 30°C,
everything below the horizontal 10 K line is
allowed. Correspondingly, at 28°C maximum office
temperature, everything would be allowed that was below the horizontal line at 12 K. We come to explain the influencing factors.
Obviously, the curves in magenta ( ), in
brown ( ), and in cyan ( ) belong to
6.6 K air temperature rise within a switched-on rack mount drawer unit, while the curves in red ( ), in green ( ), and in
blue ( ) belong to 3.3 K.
On the other hand, the curves in magenta ( ) and
in red ( ) belong to 3 kW
nominal warm water radiator heat power, the curves in brown
( ) and
in green ( ) to 6 kW, and the
curves in cyan ( ) and in blue
( ) to 2×6 kW. As a reminder, the nominal warm water radiator
heat power here refers to the EN442 standard conditions 75°C / 65°C / 20°C, and a
radiator exponent of 4⁄3
has been assumed. As expected, the nominal warm water radiator heat power dominates the situation.
Besides, the air temperature rise in the rack mount drawer units should be kept at values of about
3 K. With the comparatively low power densities here,
this is not a problem at all. Seen from the point of view of the approach, working on one of the
green lines is favored, id est, with a single warm water
radiator as described above. Hence, ambient temperatures of 30°C
still allow 400 W power consumption in the
cabinet, those of 28°C already 600 W, and those of 23°C
the full kilowatt. In certain circumstances, the warm water radiator can be placed in a larger
stairwell or basement, such that outsiders do not get access to it, or, if the space conditions permit
that, two 6 kW models can too be connected in series.
With two 6 kW models in series, even 800 W are still possible at 30°C in the environment. It should also be mentioned that the
air-water heat exchanger should possess a certain minimum
value with regard to its heat capacity to cushion impacts caused by cooling water circuit disasters.
It should further be noted that temporal temperature changes should not exceed a rate of 20 K/h as long as massive hard disk accesses occur. Of
course, this last condition can also most easily be fulfilled with two 6 kW models in series.
3. Commercial aspects
There is no doubt that there are customers for the solution presented. The problems are everywhere
the same and to rely on the 19-inch format is as consequent as
reasonable. What the experience, however, quite clearly shows is that, outside of data centers, the
usability of 19-inch data center equipment is rather limited,
and that 19-inch equipment, which is intended for customers
outside of data centers, does not meet the needs the envisaged target customer group really has.
Against such a backdrop, we want to offer an energetically reasonable one-cabinet data center with extremely low
maintenance requirements, and we want to have commercial success with it. Such a one-cabinet data center is therefore certainly not
a product for which the components can easily be gathered together. Such a one-cabinet data center is a product consisting of different,
partly multiply existing components, of which at least six have to be specially manufactured here.
These six components, which have to be specially manufactured, are the air-water heat exchanger, the pump
casing, the mentioned Reverse Flow Blockers, the
special rack chassis, the small control device, and the optional anti-drumming kit to improve
sound absorption. Of course, as in the case of the air-cooled
variant, the devil is also in the detail here. Two prototypes should soon be built. This requires funds
that Seidl Technologies alone does not have. When it comes to prototype construction, our aim is not
only to ensure that the design of the components to be manufactured turns out suitable and cost-effective, but also to permit the separate marketability of at
least the air-water heat exchanger,
the special rack chassis, and the Reverse
Flow Blockers. Perhaps even the pump casing is salable if it is offered together with air-water heat exchanger, one will see. For the special rack
chassis, also data centers are seen as a target customer group if they are going to eliminate their
messy corners. Finally, the competitors, too, which at the moment are, of course, not yet existing,
could be considered as a target customer group for all the components to be manufactured.
In summary, our main objective here is to find a highly productive metal construction company with
certain qualifications in the electronic and electrical engineering sector, which has the will and
the strength to tackle the outlined project together with us. How the cooperation in this case actually
turns out will surely be subject of negotiations. Alternatively, if you think that we have a good
thing on here, there is always the possibility to push the prototyping through direct financial
contributions. Just contact us.
A solid way to prepare contacts with us is the one via the . Please forgive
the circumstance that it is still seen that this form was originally only intended to prepare
business relations with customers. As a potential partner you will surely expand the part
Cooperation in this form.
4. About this page
This page has been online since January the 9th, 2017.
The present release of this page is the
version 4 of October the 12th, 2024.
|
|