An Investigation into the Use of Hybrid Solar Power and Cloud
Service Solutions for 24/7 Computing
Emmanuel Kayode Akinshola Ogunshile
Department of Computer Science, University of the West of England, Bristol, U.K.
Keywords: Cloud Services, Solar Energy, Energy Storage, Digital Systems.
Abstract: As the human race demands more from computing, the national grids of nations around the world
subsequently have to burn additional fossil fuels to meet increased power requirements. The aim of this
paper is to investigate ways in which an organisation could reduce its operational costs and therefore be
greener through the implementation of either a complete solar solution or a more hybrid mix with cloud
computing thrown in. Through the creation of a hypothetical UK based SME we compared solar technology
currently in the market in order to understand not only the total investment required but also just how
efficient solar technology is, or perhaps is not. We also investigated comparable technology from the three
cloud providers (Microsoft, Amazon and Google) to discover whether replacing on-premise hardware with
that available in data centres would be more cost-effective than full solar solution or reduce the total amount
of solar technology required. Having conducted the research, we found that solar technology is in no way an
effective solution for the total replacement of power from the national grid, it can be very pricey to
implement especially on the scale of always on computing and is easily affected by the elements-which
given the UK as a location is not ideal. It was also discovered that cloud computing is in no way as
affordable as it is perhaps made out to be but has the benefits of being considered a) an operational
expenditure, b) fully maintained and; c) fully flexible, these all being reasons which help a growing SME
expand down the line without unnecessary hardware outlay. Our final recommendations provide a fair cost
comparison over the total expected payback period for the solar setup of installing a solar solution to power
the entire on-premise systems and simply having a hybrid of both solar and cloud.
It was Guntenburg’s modification to the chinese
print press technique that changed the industry
forever with his new approach resulting in over half
a million books entering circulation by the end of
16th century (Whipps, 2008). Fast-forwarding to
modern day and machines of similar technique now
form an industry worth over a staggering $640
billion USD a year (Aleyant, 2015). Whilst the
industry is clearly established you could argue that it
faces an ever increasing threat from the invention of
the 20
century, the internet. With prestigious
organisations such as the Oxford English Dictionary
moving to online-only production (Wang, 2014), this
makes the threat evermore apparent. The Guardian,
on the other hand, disagrees and claims that more
than 80% of respondants still prefer to “consume
articles via print” (Jamieson, 2010). Furthermore,
many printers such as Exaprint have adapted to
provide online-only ordering and production
services which, they claim, result in upwards of 30%
growth year on year (Exaprint, 2015).
In 2005 forty-five trillion pages were printed in
the United States alone (Aleyant, 2015) therefore
creating a demand for machinery that can offer both
increased times of operation and higher production
volume. With many print presses now running 24-
hour operations their business owners are facing an
operational strain in the form of electrical costs.
In this paper we assume the identity of
hypothetical company Bristol Solar whom have
recently entered the bespoke solar market. Their
customer CMYK is looking for a solution to reduce
the operational costs of their on-premise I.T.
hardware through the introduction of solar
technology. Bristol Solars specialise in the
deployment of cloud solutions in order to reduce
overall energy consumption. This paper is organised
as follows: Section 2 introduces and discusses the
customer (CYMK) and their existing infrastructure
Ogunshile, E.
An Investigation into the Use of Hybrid Solar Power and Cloud Service Solutions for 24/7 Computing.
DOI: 10.5220/0006380007430754
In Proceedings of the 7th Inter national Conference on Cloud Computing and Services Science (CLOSER 2017), pages 715-726
ISBN: 978-989-758-243-1
Copyright © 2017 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
and section 3 looks at the power consumption of that
infrastructure. Section 4 look in detail at the types of
solar panel technology and how environmental
factors can affect them before finally concluding
which one would suit CMYK. Section 5 presents
information on cloud-computing offerings such as
Infrastructure as a Service (IaaS) etc. and discusses
how CMYK might introduce such offerings and
Section 6 presents the conclusion and cost analysis
of purely converting to solar or employing a hybrid
cloud and solar solution to reduce overall energy
CMYK are a small print organisation with no more
than 30 employees between them yet they each
contribute to the completion of more than 1 million
customer orders per year. Whilst the majority of
employees are warehouse floor staff CMYK also has
your typical back off staff who make use of your
typical Windows operating system to operate the
overall business, from finance to pre-production and
management. In addition, CMYK also operate an on
premise I.T. environment designed specifically for
the storage of customer artwork as submitted via
their online ordering system. You can find a detailed
outline of CMYK’s IT infrastructure in section 2.2
of this paper.
2.1 Recent Capital Investment
Ever since CMYK have moved to online sales they
have seen orders sky rocket which has lead investors
to require that CMYK increase their profibability by
reducing operational expenditure. CMYK are
therefore seeking an energy efficient solar panel
solution to provide 24-hour continuous power to the
in-house server room and have recently replaced all
servers with those that provide more, for less power.
CMYK has done away with typical
desktops in favour of an on-premise, server-driven,
thinclient approach.
2.2 Existing Infrastructure & its
CMYK use a combination of Hewlett-Packard
systems and DELL servers each running
virtualisation software such that they operate the
main website, the online ordering system and your
typical Windows Server operating systems. In
quarter one of 2015 alone CMYK had stored
terabytes of historical customer media alone. The
majority of hardware has recently been
introduced/replaced as part of capital investment, see
2.1. In addition to general power efficiency afforded
by solar technology, CMYK are also looking for a
recommendation as to reducing on premise hardware
operational expenditure and have asked for advice
on cloud offerings which, as it happens, Bristol Solar
specialise in. The servers/equipment CMYK
currently use are as follows:
1. HP ProLiant ML110 Gen9 Server, Xeon E5-
2600, 60GB Memory. Used to operate the
majority of print system software and operations
in addition to the hosting of the company’s
websites and order systems.
2. HP ProLiant ML110 Gen9 Server, Xeon E5-
2600, 180GB memory. Used to operate the
compaies terminal services computing
virtulisation. CMYK make use of HP t520 thin
client machines to connect to this server.
3. Dell PowerEdge T110 II compact tower server,
Xeon E3-1200, 20GB Memory. Used to operate
Windows Servers, Active Directory, Print
Servers and general back office systems.
4. HP 3PAR StoreServer File Controller.
5. HP 3PAR StoreServ 8450 Storage unit. In place
to service CMYK’s future storage needs.
2.3 Customer Location
For the purposes of working out solar panel
efficiency, CMYK are located in Avonmouth just off
the M5 in Bristol and have a purpose build building
that is complete with a flat roof surface, perfectly
suited for the installation of solar panels.
The aforementioned infrastructure has the following
power consumption requirements expressed in
HP ProLiant ML110 Gen9 Servers
0.55 KWh – When looking at the data sheet
provided by HP for their Proliant ML110 Gen9
series server they listed the power requirement for
one individual unit in standard use at 350W. With
that said, no standard power supply unit can
typically provide 100% efficiency and therefore HP
claim 8% power is lost during AC or DC conversion.
CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science
With this in mind, the true power draw for this
machine must be at least 381W. Converting from W
to KWh is as simple as converting from litre to
millilitre. See this equation:
The power P in kilowatts (kW) is equal to the power
W in Watts divided by 1000. Expressed as,
Given CMYK operate two of these systems the total
power is therefore 0.762Kw.
Dell PowerEdge T110 II compact tower server
2.11KWh – According to the hardware datasheet we
know that the PowerEdge T110 II uses between 90 –
264 Voltage AC power depending on use. We also
know that the machine, when converting the power
for its own use operates at 80% efficiency and based
on a UK supply of 240V understand the unit to
require 10 Amp supply. Knowing all of this, it is
possible to convert to KWh using the following
math equation.
The power P in kilowatts (kW) is equal to the power
factor PF times the current I in amps (A), multiplied
by the voltage V in volts (V) divided by 1000.
Expressed as,
HP 3PAR StoreServer File Controller
1.6KWh – The 3PAR StoreServer data sheet
highlights that this controller system makes use of 2
x 800W power supply units encased within the
chassis. This time, however, unlike the HP ProLiant
ML110 we are not provided with any information on
the expected power draw as this varies massively
dependant on the amount of storage devices you
typically install in to the controller. For the purposes
of this research we shall presume the controller
draws with a similar efficiency to the ProLiant
server above and only uses the second PSU as a
backup, using say 10% when idle. Applying the
logic from the calculation of the ProLiant, we arrive
at 736 Watts (main) and 8 Watts (idle). When added
together we get 816 Watts of consumption, espressed
as 0.816 KWh.
HP 3PAR StoreServ 8450 Storage
1.11KWh - It is with this unit where things become a
little more complex. With each additional plug-and-
play hardware device, such as a Fibre Channel card,
comes additional wattage power requirements. In it
is standard configuration the unit can share power
with the File Controller. The unit on its own with no
drives draws 803W of power. 24 drives of a
reasonable capacity (2TB each) draw 13.1 Watts and
a 10Gb/s network card draws 5.71 Watts. Adding
these together we see a total requirement of
1123.11W. Otherwise known as 1.12KWh.
Total Annual KWh Requirements
Based on the above calculations, and the fact CMYK
require their I.T. infrastructure to operate 24 hours a
day, 7 days a week, the total kilowatt power
requirements over a 365-day period expressed as TP
(Total Power) is,
According to statistics published by the (Energy
Saving Trust, 2015), the total average cost per KWh
of electricity in England is 14.05 pence and using
this figure would indicate that it costs CMYK a total
of £5,907 per annum, or £492 per calendar month to
run its I.T. server infrastructure.
In the year 2015 solar panel technology is more
widespread than even half a decade ago and
America’s Fortune Magazine confirms they are not
“just for rich home owners anymore” (Fehrenbacher,
2015). As adoption levels increase, with the UK now
within the global top ten (Solar Trade Association,
2014), so does the differing number of technologies
which satisfy the same goal. In this section of the
paper we take a look at the different types of solar
technology and make a recommendation as to the
most suitied for CMYK.
4.1 Differences in Technology
There are many different types of solar panel
technology in the world but in the mainstream
market there are typically three different types of
panel in use, each with its differing uses. We explore
each of them here with information made available
with thanks to Alternative Technology Association
(herin ATA) and the Energy Informative (herin EI).
An Investigation into the Use of Hybrid Solar Power and Cloud Service Solutions for 24/7 Computing
Figure 1: Monocrystalline Silicon Solar Cells (Maehlum,
According to the ATA, Monocrystalline panels are
cut in thin wafers from a single piece of silicon and
all associated conductive cabling is integrated in the
surface of the solar panels themselves. This
therefore produces a very efficient solar panel
which, at the time of writing, EI claim to be the most
effeicent in the entirety of the U.S. market.
EI tell us “Monocrystalline solar panels produce
up to four times the amount of electricity as thin-
film solar panels” and in April 2013 the solar
company SunPower announced their new X-Series
panel. This new panel could produce an average of
21.5% efficiency – one of the highest ever seen from
a panel of this type.
At the time of writing, panels of this type are
widely manufactured by companies such as LG,
Sharp among others. What is more, many of the
aforementioned have faith in their technology by
providing warranties as long as 25 years.
The ATA say that whilst these panels may be
very efficient they certainly come at price but
continue to say that whilst initially expensive they
would pay for themselves within 5 years in most
For all of their positives though, this type of
panel certainly has one key negative. EI claim that
should only one unit be covered by shade, snow or
anything else it could have an adverse affect on the
energy production rate of the entire collection of
units. However, EI continue to say this affect can be
reduced with additional equipement, a micro-verter,
albeit at increased cost.
EI tell us that Polycrystalline cells often cost less
as a result of a cheaper production process and
according to ATA this panel type is made from
separate wafters of silicon as opposed to one entire
cut, resulting in less chemical waste.
Figure 2: Polycrystalline Silicon Solar Cells.
ATA claim that this type of panel has a higher space
efficiency as a result of its square moulded
appareance. By contrast, Monocrystalline are often
more circular and you are therefore unable to house
as many cells within a single unit. Whilst EI agree,
they do say that Polycrystalline’s depedenance on
several silicon wafers as opposed to a single cut of
Monocrystalline significantly reduces the output
effieciency by as much as 8%. Therefore, EI
continue by saying you would require many more
panels in order to compensate for difference you
could otherwise achieve with Monocrystalline units.
It would seem from this research that the
Polycrystalline units, whilst more affordbale, are a
lot less effecient and are perhaps not suitable for a
commercial environment. EI claim that most
concerns they bring up would not really affect
someone in a home, but may be the deal breakers for
Amorphous / Thin Film
Figure 3: Thin-Film Solar Cells.
This type of panel is produced by affixing thin strips
of solar technology on to a substrate such as glass or
metal and according to EI this makes it perfectly
suited to production en-masse. What is more, this
type of panel is not impacted by heat and shading
CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science
issues experienced by Polycrystalline and
Monocrystalline respectively.
EI say that of the three panel types thin film is
the least economical with figures around 7-13%
mark. This would arguably see it much less suited to
an at-home environment. However, with that said,
thin-film is very flexible and as a result can be
placed over a much larger surface area and even in
places traditional panels would otherwise not be able
to go.
Example uses of Thin-Film Solar Cells include
devices such as watches and calculators all the way
up to large-scale solar farms, as pictured.
Recommendation at this Point
Based on CMYK’s initial requirements and flat roof,
Bristol Solar recommends Monocrystaline panels
due to the higher likelihood of seeing a return linked
to their higher efficiency ratings. Whilst they may
initially cost more to install, the consistently higher
energy production rates will ensure CMYK see a
faster return.
4.2 Efficiency Ratings
Understanding which technology is best suited to
CMYK is one thing, but are the panel types truly
efficient enough for the customer’s requirements?
Wind&Sun, a UK distributor of solar panels state
that “Modules nearly always produce less than their
rated peak power in real-life conditions”
highlighting you can not simply rely on a
manufacturers specification sheet alone. So how
exactly do you calculate the true efficiency of a solar
panel? Answering this question will help Bristol
Solar in recommending a suitable panel solution.
Using calculations on the Civic Solar’s website
we can immediately determine the extent to which
solar technology lacks compared with fossil fuel.
The example claims 14.5% efficiency and this
highlights the very important need to research and
shop around to ensure CMYK buy with long-term
efficiency in mind rather than the figure on the
The 14.5% system in question has a total input of
1.6KWh from the sun (based on reversing the
calculation), but loses a staggering 85.5% of this in
generating and converting its output 0.24KWh. This
is without taking other facts such as hours of
sunlight, panel elevation and shade in to
consideration. We look at these factors in 4.3.
The efficiency calculation is as follows, where E
is efficiency, W is the manufacturers rating in Watts
and SA equals the surface area of the solar panel in
4.3 Efficiency Factors
Whilst sunlight is one of the only natural resources
that keeps on giving day in, day out, rain, wind or
shine, constant energy generation can not be as
simple as installing it and there must be factors that
prevent maximum energy generation. In this section
we take a look at factors which impact maximum
solar panel efficiency to ensure we can product
installation recommendations.
UK Weather
The existence of the word ‘solar in solar panelling
is indicative of a relationship with the sun.
Therefore, deductive reasoning can conclude that
solar panels require the sun in order to function.
Whilst an overcast day might not be very sunny, the
sun is still present and therefore panels do
to function, although at what we presume would be a
less efficient rate of energy production.
We felt it rather important to see just how much
sunlight Avonmouth, the location of the CMYK
facility, received on an annual basis. According to
the Met Office the nearest official weather station to
Avounmouth is Filton, a total of 8 miles away by
road. It is worth noting here that whilst this paper
was written in the years 2015/16, the Met Office
data relates to the 29-year period ending in 2010.
With data spanning such a wide time period it is
more than likely, almost certain infact, that any
averages created from it will vary significantly from
the annual sunlight hours in 2010 alone, or 2015 for
that matter. This is most likely due to changes in
both climate and the angle of the sun since records
began. Nevertheless, the fact records exist in the first
place enables us to improve the findings of our
research investigation.
In Table 1, we can see the total number of
sunlight hours received at Filton weather station per
year is an average of 1627. We can calculate this as
a rounded average of 4 hours of direct sunlight per
day. When comparing the sunlight of Avonmouth
with that of a mainland European city, Barcelona,
there will be a distinct difference. Using data from we could produce the table
found in the adjacent column.
An Investigation into the Use of Hybrid Solar Power and Cloud Service Solutions for 24/7 Computing
Table 1: A table to show the average number of sunlight
hours in the Avonmouth area on a month by month basis.
Month Sunshine (Hours)
Janurary 58.5
February 74.8
March 112.7
April 170.8
May 199.6
June 214.7
July 217.7
August 201.8
September 149.9
October 104.8
November 69.1
December 52.7
Annual 1627.0
Table 2: A table to show the average number of sunlight
hours in the Barcelona area on a month by month basis.
Month Sunshine (Hours)
Janurary 155
February 140
March 186
April 210
May 248
June 270
July 310
August 279
September 210
October 186
November 155
December 155
Annual 2499
You can see that Barcelona receives an annual
average of 2499 sunlight hours. As a rounded daily
average this can be expressed as 7 hours of direct
sunlight. That is a 75% increase over Avonmouth!
With such a stark difference in this figure we can
conclusively say that solar panel technologies will
already be at a significant disadvantage in the United
Kingdom on sunlight hours alone. This is essential
knowledge as with most solar panels producing
energy with at most a 21.5% efficiency rate, every
single hour of sunlight counts if the final solution is
to produce a worthwhile return on investment.
How and why Does Shade Impact a Panel?
To provide the best solution for CMYK we need to
understand why shade impacts a panel and how it
actually does it. In this section shade refers to any time
where the sunlight is not directly upon the panel
surface. This can be caused by overcast skies, or any
object overshadowing a solar panel unit, such as a tree -
although, one would hope a panel never be installed in
close promximiity to a tree in the first place.
There are two definitions for shade on a panel:
soft and hard sources. A soft source is something
such as the shadowing from tree leaves, roof vent or
chimney. Further, a soft source may not be present
all of the time and is usually caused by the suns
position in the sky. A hard source on the other hand
is deemed to be caused by something physically in
contact with the solar panel unit. For example, a tree
branch, a blanket or perhaps an animal/bird sat on
top of it. The same paper continues to say that in
tests of an array of solar panels, simply one unit
being partially hard-shaded drops the power
generating capability of that unit by half and a
completely covered unit would produce no output
whatsoever and actually end up costing money.
Figure 4: A graphic produced to demonstrate the impact
soft-shading can have on an array of solar panels.
A typical installation of solar panels sees units
arranged in series and parallel. According to
research, connecting them in this way enables them
to produce an output voltage and current typical to
most common applications - something a single
panel would not be able to achieve on its own.
During our research, it was observed that whilst
linking panels produces a positive result, it can be
adversely affected in times of shade.
Further, when a cell is shaded, the number of
electrons it can pump from one side to the other
drops. This would not be a huge problem for a single
cell although in most environments units are
interconnected to form arrays. It is in these
situations, that shade has an impact on the amount of
power a units’ neighbour can subsequently produce
too. The output of the shaded unit then becomes the
input of its neighbour. You can see an example in
Figure 4. A completely shaded unit can impact the
efficiency of the entire array by as much as 50%.
CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science
Figure 5: An example solar array with both bypass and
blocking diodes present.
In a separate paper on this topic, H. Patel et al.
produce a software simulation of solar panel
shading. Through various simulations they
discovered that the use of bypass and blocking
diodes, within an array significantly improves its
performance (see Figure 5). The presence of such
diodes sees that units with the poorest performance
are simply bypassed therefore removing the
opportunity for energy loss. Whilst this is great in
theory, MidSummer are quick to say that there are
still many cells within a poorly performining unit
that collect energy exactly as expected. It would be a
shame to simply discount it, they say, and comment
that this is the main reason manufacturers place
diodes at a cell level. As an example let us look at
the 3rd set of shaded modules in Figure 4. A
manufacturer may choose to install bypass diodes at
the tip of the shaded segment thus meaning all
energy generated from cells 16 to 31 simply enter
above the shade of cells 31 to 50. MidSummer
suggest three solutions for solar shade:
1. Install as normal and have poor performance.
2. Act for a smaller array, only installing panels in
areas with consistent sunshine. The cost saved
can be invested in higher effeciency solar panels.
3. Split an array in to smaller arrays by installing an
inverter on each section so to collect and
consume or store the generated energy.
These three solutions are all valid. Which to use
however would depend on the owner of the new
solar panels. In the case of CMYK it would make
sense to follow the latter option as if shade impacted
energy production significantly then it would defeat
the objective of having them in the first place as they
would need to draw from the grid in order to power
their servers. MidSummer say that splitting arrays
costs significantly more as inveters and diodes do
not come at a cheap price. But, it would be a fair to
say you get what you pay for.
Elevation of the sun
With a new understanding of how solar energy is
impacted by shade we felt it important to continue
research in to external factors. The level of shade a
solar panel could experience is most likely itself
impacted by the evevation of the sun. In this section
we will research the elevation of the sun over the
CMYK building and how elevation affects a solar
panel. It is with thanks to the University of
Oregon’s Solar Radiation Monitoring Laboratory,
that I was able to generate a sun elevation graph
using the latitude and longtitude of Avonmouth,
given by Google Maps as 51 degrees, 30 minutes
and 11.9 seconds North and 2 degrees, 41 minutes
and 51.8 seconds West. See Figure 6 for the
resulting solar representation.
Whilst on first glance this graph appears to be
confusing you begin to realise its value rather
quickly once you understand that the y axis is the
suns elevation in the sky and the x axis is the
location of the sun in the sky from east to west.
Furthermore, each red marking indicates an hour of
time and the blue markings indicate the suns journey
through the sky in a given period of time as per the
indicated date. With this new understanding we can
use the data to see that in the height of an
Avonmouth Winter (December) the suns typical
elevations are:
Sunrise is ~8:30am at elevation 1.5°
Mid-Morning is at elevation 10°
Mid-Day is at elevation 12.5°
Mid-Afternoon is at elevation 4.5°
Sunset is ~3:45pm at elevation 1°
When comparing this to the height of Summer
(June) we see that:
Sunrise is ~4am at elevation 2°
Mid-Morning is at elevation 52°
Mid-Day is at elevation 62°
Mid-Afternoon is at elevation 46°
Sunset is ~8pm at elevation 2°
There is of course a huge degree of variance in this
Fig. 6. A graph of sun elevation in the
Avonmouth area. Generated using a tool from the
Universiy of Oregon.
from month to month but in picking out Winter and
Summer we can demonstrate, quite clearly, that solar
panels will be significantly affectled in Winter. This
poses the question, where is the best place to install
a solar panel, given this variance in the location of
the sun.
Charles Landau believes he has the answer and
through his research set out to establish the best
installation location with a specific focus on angle of
attack. He tells us that “Solar panels should always
face true south if you are in the northern hemisphere,
or true north if you are in the southern hemisphere”.
An Investigation into the Use of Hybrid Solar Power and Cloud Service Solutions for 24/7 Computing
Figure 6.
He claims that doing this ensures every panel
receieves exposure throughout the entire day, as
opposed to only receiving it in the morning if
installed facing eastwards, for example. On it is own
this information is not overly helpful because even
panels installed on flat roofs may be rotated to face
north/south. Thankfully, Landau has this base
covered too and tells us that panels should be
installed at your latitude plus 15° in winter and
minus 15° in summer months. He claims that
adjusting your panels twice a year provides “a
meaningful boost in energy” especially when the sun
is typically lower in the sky.
Armed with Landau’s claims we would therefore
say that effective power generation at Avonmouth
would require all panels to face true south at a pitch
of 66° in winter or 36° in summer. The dates of
change should be March 30th for summer and
September 12th for Winter. Doing this will see
panels generate 74% optimum power. When faced
with criticism over the accuracy of his calculations
Landau states “we are considering the whole day,
not just noon. In the morning and evening, the sun
moves lower in the sky and also further north (if you
are in the northern hemisphere). It is necessary to tilt
less to the south (or more to the north) to collect that
sunlight”. Whilst this is a very valid response
Landau’s admits that should the installation location
be blocked by trees or other buildings etc, then other
calculations will need to be considered. However,
for the purposes of this research paper we will not
discuss those here.
Recommendation at this Point
With all of the findings to this point we can say that
CMYK most effective solution should continue to
consider Monocrystaline panels. Any panels
installed should be both south facing and angled to
calculate for changes in the suns location. What is
more, in order that CMYK can account for poor
performance from shaded panels they should
separate panels in to arrays and make use of bypass
and blocking diodes. In section 5 we will look at a
selection of Monocrystaline panels and discuss
which would be most suitable to CMYK.
5.1 How Much Does It Cost?
Perhaps the most important question here is how
CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science
much does it cost an organisation to move to the
cloud? Well, we have already established in section
7.4 bullet point 3 that there is zero capital
expenditure to move to the cloud – so that is a great
plus for CMYK. Admitedly, CMYK have already
spent a significant amount of money on new
equipment, but even so we shall investigate the
costings involved here. As established in section 7.3,
we compare the cost accross Microsoft, Amazon and
Google where possible.
When deciding which services would suit
CMYK we take a look at their existing hardware
with a specific focus on those which consume the
most energy both now and in the future. From this
we can see that costs could potentially be reduced by
replacing both the Dell server and HP Storage
Arrays. Considering the Dell is a physical machine
offering Windows Server features a PaaS solution
would be approrpaite here. For the Storage Array we
shall look at STaaS.
IaaS – Replacing the Dell
Given that the CMYK have their Dell server
configured with 20GB of memory we shall look at
the next best comparison here to provide at least
20GB as a minimum from each of the providers. The
prices will also factor in a 365-day operation.
Microsoft Azure
Microsoft offer their IaaS Virtual Machine in a
variety of pre-configured plans. Each plan differs
in processing power and memory allocated to the
VM. Based on the Dell’s initial configuration the
A6 plan is the next most suitable comparison.
Offering 4 processor corses, 28GB memory and
285GB of inclusive storage the hourly pay-as-
you-go rate is £0.4155. The total cost over a 365-
day period is therefore £3,639.78. Whilst 28GB
is 8GB more than currently available, this is
close enough to be considered comparable given
the matching core count.
Amazon Web Services
Amazon’s take on IaaS is called EC2 but unlike
Microsoft’s offering the EC2 pricing structure
pages can get quite complicated. Nevertheless,
given the requirements of CMYK the closest
matching service is ‘m4.2xlarge’ which offers 8
virtual CPU cores and 32GB of memory. The
hourly pay-as-you-go rate is £0.69 and the 365-
day operation is therefore £6,049 but depending
on whether CMYK require the full 8 cores and
32GB of memory there is a lower plan which
offers 4 cores and 16GB memory, with a pay-as-
you-go rate of £0.34 per-hour and an annual cost
£2,978. Whilst Azure bundles each VM with
some basic storage, Amazon does not do this so
additional storage purchases are necessary atop
of the infrastructure costs, more on how much
this costs in the STaaS heading.
Google Cloud
Similar to Amazon Web Services, none of
Google’s standard packages suit the initial needs
of CMYK like-for-like. There is a plan below,
and a plan above and prices for both are include
here. The smaller of the two plans ‘n1-standard-
4’ offers 4 cores and 15GB of memory and the
larger ‘n1-standard-8’ offers 8 cores and 30GB of
memory. The pay-as-you-go and annual pricing
for the former is £0.22 and £1927.20 respectively
whilst the latter is £0.44 and £3854. Like
Amazon, these prices do not include any bundled
disk space and therefore this would be as an
additional cost. This is discussed in more detail
under the STaaS heading.
From this comparison of providers, it becomes clear
that Microsoft’s Azure offering is the most suited to
CMYK given its close match to the specification of
the DELL server. The Azure VM is by no means the
cheapest, with one of Googles offerings being
below £2,000 but we aimed to find the most
comparable offering. In adopting the Azure product
CMYK could completely remove the requirement
for the DELL server and instead transition to
Thinclient operations resulting in an annual saving
of 18,429 Kw from the server alone removing the
need for such to be covered by solar energy.
STaaS – Replacing the Storage Arrays
CMYK plan to store all of their customers’ data on a
storage array offered by HP. Initially it would appear
CYMK’s intentions were to run the storage array
rather empty with only a few drives in operation.
Given the power draw of the drive controllers the
company could offload storage to the cloud and
prevent any future increase in energy demands as
their usage increases. Going back to the research in
point 7.4 we know that cloud services are rather
flexible so CMYK could simply expand their storage
requirements as needed and later shrink them. As for
IaaS we shall compare storage services across all
three providers.
Microsoft Azure
Microsoft describes their storage offering as
“durable, highly available and massively
scalable” and offers its service with fees based
on the amount of space you truly use, not what
you initially request. The price list is structured
as £0.0147 per GB for the first terabyte per
month followed by £0.0145 per GB for the next
An Investigation into the Use of Hybrid Solar Power and Cloud Service Solutions for 24/7 Computing
49 terrabytes per month. So, given CMYK’s
current 48TB usage they would be looking at
£14.70 for the first 1TB and £681.50 for the next
47TB meaning CMYK will be charged £696.20
per month for their current storage requirements
or £8,354 for the year.
Amazon Web Services
Called Simple Storage, or S3 for short, Amazon
provides three variations on its’ storage product.
Standard, Infrequent and Archive. The former is
a typical storage similar to Azure’s, the latter is
as the name suggests, data archiving and
infrequenty is similar to Standard although
provides lower pricing on the assumption data
will only be accessed infrequently. Given the
requirement of CMYK the Standard offering is
most suitable. Pricing is fairly similar to
Microsoft’s Azure. The first 1TB is priced at
£0.021 per GB with the subsequent 49TB
charged at £0.0206 per GB. It is clear that
Microsoft’s Azure offering comes out on top here
where price is concerned. CMYKs 48TB would
cost them £21 for the first 1TB followed by
£968.20 making a total of £989.20 per month or
£11,870 per annum. This is a 42% premium over
Microsoft’s Azure offering which is rather
Google Cloud
Google’s cloud storage offering somewhat
imitates the exact offering of Amazon in the
sense they offer Standard storage, Infrequent
storage marketed as DRA and Archive marketed
as Nearline. Google also claim they offer “low
cost” storage with no teir pricing. Google’s
Standard offering comes in at a flat rate of 2p per
GB resulting in 48TB costing a total of £960 a
month or £11,520 per annum.
With the comparisons above we can tell that
Microsoft’s Azure storage appears the most
affordable with continued discounts available based
on the amount of storage used thanks to its tiered
offering. It is worth noting that one area that has not
been researched is network traffic. The above cloud
providers each charge for the movement of data
from their data centres across the internet, referred to
as Bandwidth in IaaS terms or Egress when related
to STaaS. Such charges could increase the overall
cost of one provider against another although such
cost investigations are beyond the scope of this
Based on CMYK requiring 42,048KW of energy for
a round-the-clock operation Bristol Solar would in
the first instance recommend that the business
reassesses to what extend they would require the
complete 48TB of storage space and based on the
response would recommend one of two possible
outcomes as described under the solution one and
two headings below.
The Panels
We established in section 4.1 of this paper that
Monocrystaline panels provide the edge over
competing technology due to their typically higher
energy ratings and subsequent likelihood for a return
on investment. When it came to the panels
themselves we concluded in section 5 that Sharp’s
ND-F4Q300 offered the most output at 300 Watts
with each unit costing £244+VAT. Section 4.2 saw
us learn that panels should be installed in a southerly
direction in smaller arrays each equipped with their
own bypass and blocking diodes such to attain the
best possible energy generation potential.
Solution One – Keeping full storage & No Cloud
Should CMYK decide to keep the storage hosted
locally then the final solution will need to be able to
provide enough energy for the annual requirement of
42,048KW. This of course includes overnight
power. Based on this CMYK would need the
following (inc. tax):
141 Sharp ND-F4Q300 Panels (£41,284.80).
Applicable mount brackets, etc. (N/A)
8 Lead Acid Batteries (£1,536)
Applicable convertors etc.
Installation Labor
The batteries and solar panels will cost a total of
£42,820.80. Based on CMYK’s £5,907 electricity
bill this would take 7 years and 3 months to pay for
itself in savings. As above, £42k is the minimum
cost for this solution. It is highly likely that
additional grid power will also be required should
panels drop below the marketed efficiency.
Solution Two – Using only the Required Storage
If CMYK decide that infact 48TB of data storage
capacity is too much for them then they could switch
to a cloud service provider for the majority of their
storage needs thus removing the requirements for the
HP storage controller and array. This solution also
factors in the removal of the least energy efficienct
CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science
server, the DELL. We established in section 7.4 that
cloud would afford CMYK the flexibility to expand
storage and computing power as required and that
disaster recovery could also be factored in too, if
required. Simply removing the above mentioned
devices from CMYK will see a deduction in their
energy bill of 35,372.88KW or in monetary terms a
saving of £4,969.88 meaning their overall annual bill
would then be £938. Given the new consumption of
6,775.12KW CMYK would then only require the
following (inc Tax):
23 Sharp ND-F4Q300 Panels (£6734.40)
Applicable mount brackets, etc. (N/A)
2 Lead Acid Batteries (min value £1500)
Applicable convertors etc.
Installation Labour
The batteries and solar panels will cost a total of
£8,234.40 and based on this solution being designed
to replace £4,969.88 of electricity per year then the
equipment can be seen to pay for itself within 1 year
and 8 months. Costs that are yet to be considered are
those involved in the operation of the new cloud
environment which we established in section 7.4
would cost £5,556.18 per annum with 10TB of
storage from the get go. The new annual cost of the
CMYK I.T. environment would come in at
£6,494.18 per annum after considering the existing
£938 energy cost.
Solution Comparison
Here is the monetary comparison:
Solution One:
o First Year:
£42,820.80+ Hardware
o Future Years:
Depends on any
additional grid power
requirements as on
premise hardware out
grows solar setup.
Solution Two:
o First Year:
£8,234.40+ Hardware
£938 Grid Power
£5,556.18 Cloud Services
o Future Years:
£938 Grid Power
£5,556.18 minimum cost
for cloud services. This
will increase alongside
CMYK’s consumption of
cloud services.
Figure. 9: A graph to compare both the cost of investment
and future costs associated with an all solar solution and
the hybrid solar and cloud offering.
It is worth noting at this point that the first solution
is presuming electrical requirements are not going to
change and that each panel will produce 300W of
energy. This is highly unlikely in the British Isles.
Solution Two is presuming the same, but each panel
is only there to service a small KW requirement by
comparison and the additional electricity charges
would likely be minimal should the grid need to be
called for. However unlike solution one, solution
two provides elasticity from the cloud perspective
and incorporates 10TB of disk storage from the get-
go yet should CMYK uses less then they could bring
this cost down significantly. Cloud computing with
its significantly lower setup costs can also be seen as
a form of operational expenditure (OpEx) which has
the potential to offer tax advantages to CMYK but
those will not be discussed here.
Figure 9 highlights the differing cost of each
solution and it is immediately clear that an all solar
solution has Year 1 costs 190% higher than that
those of the hybrid offering. Although the former
sees a lower year on year cost from Year 2, this will
undoubtedly increase when the solar equipment can
no longer meet future power demands. Whilst a
hybrid setup requires less power from solar panels
for on premise hardware, it has the potential to
become expensive overtime especially when you
factor in redundancy services and bandwidth etc. so
ultimately the final choice would be down to
CMYK’s own budgets. With that said, Bristol Solar
would recommend cloud computing as it would cost
significantly less down the road, enable rapid
deployment and affords so much flexibility down the
road as the business grows.
N.B. Each solution’s’ prices include VAT but would
be subject to change once brackets, cabling and any
associated labour were to be included.
An Investigation into the Use of Hybrid Solar Power and Cloud Service Solutions for 24/7 Computing
N.B. Cloud service provider bandwidth/egress is not
factored in to the final price. As a result, any
purchase decision based on this research should
consider this in more detail.
U. Wang, "Guess who are the top 10 solar panel makers in
the world?," in Forbes, Forbes, 2014. [Online].
makers-in-the-world/. Accessed: Dec. 7, 2015.
A. Jamieson, "Oxford English Dictionary ‘will not be
printed again’," in The Telegraph,,
2010. [Online]. Available: http://
again.html. Accessed: Oct. 6, 2015.
H. Whipps, "How Gutenberg Changed the World,", 2008. [Online]. Available: http://
world.html. Accessed: Oct. 19, 2015.
Aleyant, "," in Print Is
Big. [Online]. Available:
index.html. Accessed: Oct. 19, 2015.
Exaprint, "Exaprint // About the company," in Exaprint.
[Online]. Available:
page/wexa. Accessed: Oct. 19, 2015.
Energy Saving Trust, "Our calculations," in Energy Saving
Trust, 2015. [Online]. Available: http://www.
Accessed: Nov. 2, 2015.
K. Fehrenbacher, "Solar panels: Not just for rich home
owners anymore," in Fortune Magazine Tech, Fortune,
2015. [Online]. Available:
anymore/. Accessed: Nov. 2, 2015.
Solar Trade Association, "The macroeconomic benefits for
the UK of investment in solar PV," Centre for
Economics and Business Research, London, Sep.
M. A. Maehlum, "Which solar panel type is best? Mono-,
polycrystalline or thin film?," Werner, 2015.
CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science