An Investigation into the Use of Solar Power in Cloud Computing

Environments

Emmanuel Kayode Akinshola Ogunshile

Department of Computer Science, University of the West of England, Bristol, U.K.

Keywords: Cloud Computing, Solar Power, Virtualisation.

Abstract: Cisco predict that by 2019, 86% of computing workloads will be carried out within a cloud computing

environment. This is leading to the dramatically increasing need for data centre expansion which in turn is

consuming more and more of the world’s natural resources to generate the electricity needed to power them.

This paper uses a fictitious electronics recycling company called Compucycle to investigate the feasibility

and cost of integrating solar power generation into Compucycle’s IT Infrastructure compared to completely

outsourcing it to a cloud service provider. It was discovered that a complete solar power solution was not

feasible due to the excessive costs it brought to the business. It was then decided that two out of four

proposed solutions in this paper were a good fit for the business. The first being a hybrid power solution

where a small portion of power is derived from the grid along with solar power generation. The second

being the outsourced option. The third and fourth solutions were disregarded due to the fact that one was

completely unfeasible and the other went against what Compucycle wanted to achieve.

1 INTRODUCTION

Compucycle Ltd is an electronics recycling

company based in Bristol, United Kingdom (UK).

They are accredited by a number of large technology

companies such as Dell (Dell, 2016), HP Inc (HP,

2016), Hewlett Packard Enterprise (HPE) (Hewlett

Packard Enterprise, 2016) and Lenovo (Lenovo,

2016). Compucycle primarily deal with the recycling

of computing hardware. Sustainability and

environmental conservation are very important parts

of Compucycle and they’re currently in year 3 of a 5

year process to make the business as energy efficient

as possible. The next stage in this process is to

attempt to power their IT infrastructure from

renewable energy sources. Compucycle have 3 sites

spread across the UK. The Bristol, Manchester and

Chatham sites each have a recycling plant and the

head office is located in Bristol. Compucycle have

stated that they would like to continue using Hewlett

Packard hardware as they are happy with it’s

reliability and performance and don’t see the need

for change. Due to the location of their sites which

are in developed urban environments where local

legislation prohibits the use of wind or hydroelectric

solutions, only a solar solution will be feasible.

Although being able to power their IT infrastructure

Ũs a key part of their plan Compucycle stress that it

must be cost effective. Compucycle will consider

cloud outsourcing methods as part of this IT

initiative. This paper only deals with the feasibility

of powering new server hardware and not existing

networking, computing, printing or electrical

systems.

Technology companies are already using

sustainable energy sources to power their data

centres, most notable are Google and Amazon Web

Services. In 2007 Google installed the largest

corporate solar panel installation at their Mountain

View campus (Google, 2016). The 1.9MW system

has been able to produce over 3,000,000kWh of

energy per year. Amazon Web Services is another

example (Amazon Web Services, 2014). As of April

2015 25% of the power consumed by their global

infrastructure is supplied from renewable energy

resources. Their 3 wind farms and 1 solar farm in the

United States generates more than 1.6 million MWh

per year.

The objectives of this research paper are:

 Understand different types of solar power

technologies and how they are utilised in

different environments. This can be found in

section 3 of this paper.

 Understand the concept of virtualisation and

cloud service platforms. This can be found in

Ogunshile, E.

An Investigation into the Use of Solar Power in Cloud Computing.

DOI: 10.5220/0006380107550766

In Proceedings of the 7th International Conference on Cloud Computing and Services Science (CLOSER 2017), pages 727-738

ISBN: 978-989-758-243-1

727

section 4.

 Design a cloud computing solution to replace the

existing Compucycle IT infrastructure based on

the requirements set out in section 2. The

proposed solution can be found in section 5 of

this paper.

 Determine whether or not the proposed solution

can be powered by electricity generated from

solar panels which can be found in section 5.2,

5.4 and 5.5.

The rest of this paper is organised as follows:

Section 2 states the requirements Compucycle have

for the proposed solution. Section 3 examines solar

power technologies, how they are utilised, what

factors can impact solar power generation and power

storage methods. Section 4 introduces the concept of

virtualisation. Finally, section 5 presents the

proposed solutions and recommendations in order to

determine the most appropriate solution for

Compucycle.

2 REQUIREMENTS

Compucycle are currently using a number of HP

tower servers, a central storage array for backup and

networking switches to make up their IT

Infrastructure. The storage and networking is

relatively new but the servers are reaching the end of

their lifespan. Each site currently has exactly the

same hardware with the exception of head office in

Bristol.

Compucycle Ltd have stated that they wish to

transition to a virtualised environment in order to

reduce underutilisation of resources, real estate

space and power consumption. As part of this

Compucycle want to attempt move their entire IT

infrastructure to the Bristol head office. Table 1

shows the resource requirements for their systems.

3 SOLAR POWER

In recent years the adoption of solar energy

generation has increased exponentially. The

(International Energy Agency, 2015) states that the

biggest adoption rate has been since 2010, where the

total global solar power capacity was 40,336MW.

Since then it has continued to grow year on year

when in 2014 the total global capacity is

178,391MW figure 1. In 2015 this was expected to

reach 233,000MW, at the time of writing this, that

data was not available. As shown in figure 1 a large

Table 1: Resource requirements for Compucycle systems.

Figure 1: A graphical representation of the adoption rate

of solar power. (Growth of Photovoltaics, 2016).

portion of this growth was in Europe. As of 2014 the

UK was in the top 10 countries globally having the

capacity to produce a total of 5,104 MW. This is a

relatively small amount compared to Germany who

have the capacity to produce 38,200MW.

3.1 Solar Panel Technologies

Almost 90% of the world’s solar panels are made

from some type of silicon. The main difference is

the purity of the silicon which has a direct impact on

the efficiency of the solar panel. Silicon purity is

determined by how aligned the silicone molecules

are, this is referred to as Crystalline Silicon. From

this it is possible to create a number of different

types of solar panel. Energy Informative has written

an article on the different types of solar panels, a

summary of which can been seen in section 3.1.

3.1.1 Polycrystalline

Polycrystalline based silicon solar panels have been

on the market since 1981. They have an efficiency

range of 13 – 16%. Being able to generate 255W of

electricity.

CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science

728

Advantages:

 The process to make polycrystalline solar panels

is simple and more cost effective.

 The process to make polycrystalline produces

less waste silicon compared to others.

Disadvantages

 The efficiency of polycrystalline solar panels is

considerably less than monocrystalline based

solar panels.

 Polycrystalline solar panels have a much lower

space efficiency compared to other solar panel

types.

3.1.2 Monocrystalline

Monocrystalline solar panels are made up of

cylindrical silicon ingots. Monocrystalline solar

panels are the most efficient solar panels with

having a maximum efficiency of 21.5% being able

to generate up to 345W of electricity per year.

Advantages:

 Monocrystalline solar panels are more space

efficient compared to other types of solar panels.

 They have a longer life span than other types of

solar panels.

 Monocrystalline solar panels perform better in

lower light conditions.

Disadvantages:

 They are the most expensive type of the solar

panel types.

 If the panel becomes covered with debris or

shade the panel will cease to operate.

 The production of Monocrystalline solar panels

produces a large amount of waste silicon.

 Monocrystalline solar panels are not as efficient

in warmer weather.

3.1.3 Thin-Film Solar Panels (TFSP)

Thin-Film Solar panels are relatively new compared

to Crystalline based solar panels. They are

constructed by depositing one or several layers of

solar cell material onto conductive material. TFSP

prototypes have reached an efficiency rate of 7-13%

and production models operate at 9%.

Advantages:

 Mass production of TFSP is very simple.

 They can be made flexible opening up numerous

more applications.

 They are less affected by temperature and

shading.

Disadvantages:

 They require a large amount of space.

 They don’t produce as much electricity

compared to other types of solar panels.

 They degrade a lot faster than other types of

solar panels.

It is clear from this that the Monocrystalline solar

panel is the better performing solar panel with much

higher efficiency, smaller footprint, longer lifetime

and better performance in lower light conditions.

Therefore, the recommendation for Compucycle at

this point is to use monocrystalline solar panels.

Comparison of a range of monocrystalline solar

panels can be found in 3.4.

3.2 Factors Affecting Solar Power

Generation

A report written by the NREL comparing studies of

solar panel degradation across a diverse range of

countries and environments states that the speed and

amount they degrade varies greatly on the type of

solar panel as well as the environment that it

operates in.

3.2.1 Sunshine Hours

Geographical locations can impact the ability to

generate solar power as different areas of the world

have varying amounts of sunshine hours throughout

the year. Sunshine hours need to be taken into

account when exploring solar solutions. Table 2

shows the average amount of sunlight hours over a

29 year period at the closest weather monitoring

station to the Compucycle Bristol site.

Table 2: Average sunlight hours over 29 years at the

Filton Weather station provided by the Met Office.[12].

Month Sunshine Hours

Jan 58.5

Feb 74.8

Mar 112.7

Apr 170.8

May 199.6

Jun 214.7

Jul 217.7

Aug 201.8

Sep 149.9

Oct 104.8

Nov 69.1

Dec 52.7

Annual 1627.0

When this is compared to a different geographical

location entirely such as Las Vegas, the amount of

An Investigation into the Use of Solar Power in Cloud Computing

729

sunshine hours per year is completely different.

Table 3 shows that Las Vegas has more than twice

the amount of sunshine hours as the UK. Bringing

into question the amount of power than can actually

be generated from solar technology in that particular

location.

Table 3: Average sunlight hours in Las Vegas, Nevada

provided by Weather2Travel.com.

Month Sunshine Hours

Jan 248

Feb 252

Mar 310

Apr 360

May 403

Jun 390

Jul 403

Aug 372

Sep 330

Oct 310

Nov 240

Dec 248

Annual 3866

3.2.2 Placement of Solar Panels

The sun moves across the sky at different elevations

during different periods of the year. Therefore, it is

important to understand the impact sun elevation on

solar power generation. Figure 2 is a graph that

shows the elevation of the sun at the Compucycle

Bristol head office site. This graph specifically plots

the elevation of months December to June. The

graph shows a massive variation in elevation making

it difficult to best determine at what angle to install a

solar panel. During summer the elevation is as

follows:

 Sunrise: 04:00AM at elevation 2°

 Mid-Morning: 10:00AM at elevation 53°

 Mid-Day: 12:00PM at elevation 62°

 Mid-Afternoon: 3:00PM at elevation 46°

 Sunset: 8:00PM at elevation 2°

Winter elevation is as follows:

 Sunrise: 08:00AM at elevation 1°

 Mid-Morning: 10:00AM at elevation 9°

 Mid-Day: 12:00PM at elevation 14°

 Mid-Afternoon: 3:00PM at elevation 5°

 Sunset: 3:50PM at elevation 1°

From this it is clear that if the solar panels are

installed to best capture the sunlight during summer

will capture very little or no sunlight during the

winter months.

Charles Landau states that it has always been

common practice to place solar panels facing true

south in the Northern Hemisphere and true north in

the Southern Hemisphere. This is to ensure that

sunlight always reaches solar panels throughout the

day. The elevation paths in figure 2 seem to support

this. Whilst this may be true for most cases

Professor Ralph Gottschalg from Loughborough

University claims this method is wrong and can lead

to spikes of energy at midday.

Although this may sound like good news to

someone storing energy this is in fact bad news if

they are feeding energy back to an electricity grid.

Gothschalg uses Germany in his example, a country

that has the most solar panel installations in Europe.

He states that having too many installations facing

south causes huge spikes in the electricity grid at

midday making the grid not only unstable but

forcing suppliers to give power away very cheaply

because there is so much of it.

To combat this Gothschalg recommends that the

UK follows Germany’s new policy of only installing

solar panels facing east or west in order to achieve a

smoother flow of electricity throughout the day.

Although this information is very useful it does not

give any reference on what angle to best place the

solar panels. Landau however does give guidance on

this problem.

He recommends that it is best to adjust the

elevation of panels twice a year. The elevation

angles should be the latitude of the location plus 15°

in winter and minus 15° in summer. From this it is

possible to work out the best angles for solar panel

elevation for the Compucycle Bristol site:

 Summer: 37.53°



Winter: 67.53°

3.2.3 Shading

It is clear that solar panels need direct uninterrupted

sunlight in order to perform at their highest levels.

When talking about solar panel shading it does not

only mean shadow created by physical objects.

It also extends to weather conditions such as

overcast skies. A paper written by G. Pachpande et

al studying the effects of solar shading, describes

two forms of shading:

Soft source shading: An object that is in the path

of the sunlight reducing the amount of light reaching

the solar panel. Examples are trees, birds, leaves and

rooftop chimneys. Soft sources do not permanently

stop light from reaching the panel and often cast

shadows that move across the panel with the sun.

Hard source shading: This can be described as

objects that completely prevent sunlight from

CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science

730

reaching the panels. This is commonly debris that is

resting on the panel itself such as leaves.

Figure 2: A graph plotting the elevation of the sun at the

Compucycle site in Bristol from Jun -Dec. Generated

using a tool created by the University of Oregon.

Figure 3: Example of Solar Panel Shading.

Pachpande goes on to state that when a single solar

panel cell is covered by a hard source the voltage of

that module will drop to half of it’s unshaded value

in order to protect itself. If enough cells are shaded,

then the panel will not generate any electricity at all

and will become a drain on the system. Figure 3 is a

graphic showing shading on a solar panel.

Considering the Compucycle head office in

Bristol is also a recycling site that is not surrounded

by other buildings and trees, there will be very little

shading in this location.

3.3 Methods for Utilising Solar

Technology

There are many different ways of utilising solar

energy to generate power. Some methods are for

small domestic and business implementations others

are for larger implementations built to feed power

straight into the grid. This sections explores the

different methods for utilising solar technology.

Figure 4: Diagram of a rooftop based solar panel system.

3.3.1 Conventional Rooftop

Rooftop solar panels have become more and more

prevalent in the UK in recent years. This is mainly

due to government grants and schemes where

surplus energy can be sold back to energy suppliers.

Figure 4 is a system model showing the process of

how solar energy is converted into electricity for

home use. DC current produced from the solar array

is transferred to a charge controller where it is

regulated and transferred to a battery for storage.

Either from the battery or straight from the charge

controller the DC current it transferred to an inverter

where it is converted to AC current thus suitable for

home appliance use. Power being transferred back to

the utility grid is transferred via the inverter. Figure

4 does not show this.

3.3.2 Solar Farms

Solar farms are the large scale implementation of

solar panels that generate electricity to feed straight

into the utility grid. They can range between 1 acre

and 100 acres in size and are often found in rural

areas. They are eligible for grants and discounts

from the UK government. They are currently the

most natural and eco friendly way to generate

energy. Solar farms are designed in such a way that

they can be combined with agricultural farming thus

making the maximum use of the land they use. To

put things into perspective for every 5MW solar

farm installed it can produce electricity for up to

1,515 homes. Each 5MW installation takes up

approximately 25 acres of land.

3.3.3 Concentrated Solar Power

Concentrated solar power generates electricity by

using mirrors to concentrate the sun’s energy and

convert it to high temperature heat. This heat is then

captured and converted into electricity. There are

An Investigation into the Use of Solar Power in Cloud Computing

731

three different methods for concentrating the solar

energy and capturing it:

Trough Systems: This is a large U shaped

reflective surface with a cylinder filled with oil in

the middle. This oil is heated up to 750°F which in

turn is used to boil water to create steam to power

turbines and generators. Figure 5 shows how a

trough system works.

Power Tower Systems: This method uses large

flat reflective surfaces to track the sun’s rays

throughout the day and reflect them towards a

central receiver. Again this receiver contains liquid

that’s heated to produce steam to power turbines and

generators in order to produce electricity. Figure 6

shows a power tower system.

Dish Engine Systems: This method uses a large

reflective dish to reflect sunlight to a receiver

mounted on the focal point. This dish tracks the

sunlight as it moves across the sky. The dish is

connected to an external combustion engine filled

with hydrogen gas. When the heat is transferred

from the dish to the engine the gas expands causing

the pistons inside to turn and power a crankshaft

connected to an external electricity generator. Figure

7 shows the dish used in this method, it is very

similar to a conventional satellite dish but they are in

fact much larger.

Figure 5: Trough System for Concentrating Solar Power.

Figure 6: Power Tower System for Concentrating Solar

Power.

Figure 7: Dish Engine System for Concentrating Solar

Power.

3.4 Comparison of Solar Panels

As established in section 3.1 monocrystalline solar

panels are the better performing. This section

compares several solar panels to find the best fit for

Compucycle. According to the solar panel

comparison website theecoexperts.com the top

monocrystalline solar panels are:

 SunPower X21-345-COM

 Sanyo HIT Double 195

 SunPower 327-320

 AUO SunForte PM318B00

Tables 4, 5, 6 and 7 give further specification of

these panels.

CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science

732

Table 4: Technical information on the SunPower X21-

345-COM.

SunPower X21-345-COM

Maximum Efficiency 21.5%

Maximum Power Output 345W

Solar Cells Per Panel 96

Size HxWxD (mm) 1559 x 1046 x 46

Warranty

25 Year Product and

Power Warranty

Price

Not publically

available

Table 5: Technical information on the Sanyo HIT Double

195 solar panel.

Sanyo HIT Double 195

Maximum Efficiency 20.5%

Maximum Power Output 249W

Solar Cells Per Panel 96

Size HxWxD (mm) 1351 x 898 x 60

Warranty

5 Year Product and

20 Year Power

Warranty

Price $849.00 (£507.09)

Table 6: Technical information on the SunPower 327-320

solar panel.

SunPower 327-320

Maximum Efficiency 20.4%

Maximum Power Output 327W

Solar Cells Per Panel 96

Size HxWxD (mm) 1559 x 1046 x 46

Warranty

25 Year Product and

Power Warranty

Price

Not publically

available

Table 7: Technical information on the AUO SunForte

PM318B00 solar panel.

AUO SunForte PM318B00

Maximum Efficiency 19.5%

Maximum Power Output 318W

Solar Cells Per Panel 96

Size HxWxD (mm) 1559 x 1046 x 46

Warranty

10 Year Product and

25 Year Power

Warranty

Price

Not publically

available

Notes: ecoexperts.com stated Ja Solar JAC M6PA-4

as the second best solar panel, upon further research

this is in fact an individual solar cell. Therefore, it

has not been included in this paper. SunPower solar

panel prices are not publically available as they are

not sold individually.

It is clear that the SunPower X21-345-COM best

panel with a much higher power output of 345W

compared to the others and a higher efficiency rate

of 21.5%. It offers the best warranty out of the

selection. It is recommended that Compucycle use

this panel or SunPower 327-320. There is very little

difference between these two panels and are the best

performing out of the selection. Price will have to be

accounted for but is not possible in this paper due to

prices per panel not being published by SunPower or

AUO.

3.5 Solar Battery Technologies

Kathie Zipp claims that the batteries used in solar

power systems have to operate under unusual

circumstances compared to others. They have to

cope with unstable and irregular power flows along

with heavy charging and discharging. For a

conventional battery this would cause significant

damage to the battery. Two types of battery

technology, Lead Acid and Lithium Ion are the most

common in solar power systems. Deep cycle lead

acid batteries are very similar to car batteries and

have been used for off-grid power storage for many

decades due to their low cost, reliability and

predictable performance. Solar lead acid batteries

are designed to be able to charge at very low

currents so they are able to take advantage of any

power available.

Lithium ion batteries are commonly found in

most household electrical goods such as laptops and

smartphones. However, they are applicable in solar

energy environments. They have high energy

density making the battery much lighter and smaller

than lead acid batteries, the life of the battery is

much longer and very high efficiency ratings

providing much faster charge times.

3.5.1 Tesla Powerwall

The Tesla Powerwall is a wall mounted lithium ion

battery developed by Tesla Motors. It charges using

power generated from solar panels and stores it for

later use. Each Powerwall unit has a 6.4kWh storage

capacity which is sufficient for the average home,

however it is possible use multiple batteries in a

single environment. Tesla is not only targeting the

Powerwall at home users but commercial

environments as well with a 10kWh variation of the

Powerwall called Powerpack. May 2015 Tesla

announced it would be partnering with Amazon Web

Services to trial the Powerwall technology in their

data centres with a 4.8MW installation in their US

An Investigation into the Use of Solar Power in Cloud Computing

733

West data centre. The cost of a Powerwall is $3000

and $3500 for the Powerpack. At the time of writing

that is approximately £2112 and £2464.

Here the clear choice for Compucycle would be a

Lead Acid Battery. It is a proven and reliable

technology that has been used for a number of years

to store power generated from solar panels. The

recommended battery for Compucycle is the

Sonnenschein A600 Solar OPzV Gel 3500. Each

battery is able to store up to 42 kWh of energy. Each

battery costs £1,578.78 which is half the price of the

Tesla battery and comes with over four times more

storage.

4 VIRTUALISATION

TECHNOLOGIES

The concept of Virtualisation can be defined as

using hardware and software to create the illusion

that two or more entities are present when only one

physical entity exists. There are many forms of

virtualisation, the most common computing forms

are:

Server Virtualisation: Making multiple servers run

virtually on one physical server. This provides

greater compute resource utilisation across a set of

servers for example. These can also be referred to as

Virtual Machines (VMs).

Desktop Virtualisation: Allows a user to switch

between different operating systems on the same

computer. These can also be referred to as Virtual

Machines (VMs).

Network Virtualisation: This gives the illusion that a

user is connected directly to a network even though

no physical connection exists. This is most

commonly used to create Virtual Private Networks

where users are access a network via the internet

without physically being connected to the network.

Storage Virtualisation: Similar to server

virtualisation, it gives users the ability to create

individual virtual hard disks that can spread across

several physical hard disks. Again providing greater

resource utilisation due to data not being restricted

to a physical hard drive.

4.1 Hypervisors

Virtualisation is achieved through the use of

Hypervisors. A hypervisor is a layer of software that

sits below the virtual machines and above the

hardware it is running on. They control the

communication between the virtual machines and

the hardware. Without the Hypervisor multiple

operating systems will want simultaneous control of

the same piece of hardware. There are two types of

Hypervisor, Type 1 and Type 2. Type 1 is often

referred to as a bare-metal implementation. Meaning

the Hypervisor runs directly on the compute

hardware essentially as it’s own operating system

see figure 8.

Figure 8: Type 1 Hypervisor Architecture.

Type 2 is where the hypervisor runs on a hosted

operating system. Figure 9 is an example of a Type 2

Hypervisor. Type 2 hypervisors are considered

easier to implement compared to Type 1 but can be

slower as it has to rely on the hosted operating

system to handle the communication to the

hardware. Type 2 hypervisors are considered less

secure than Type 1 due to the fact that the host

operating system could become infected with

malicious software.

Figure 9: Type 2 Hypervisor Architecture.

4.1.1 Hypervisor Features and Examples

According to Software Insider the typical features in

a Hypervisor are: VM creation, VM cloning, VM

backup/restore, VM snapshots, live migration,

dynamic resource allocation, VM migration, storage

migration, performance reports, virtual firewalls,

failover and thin provisioning.

4.1.1.1 VMware vSphere

vSphere is a data centre management and

CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science

734

virtualisation application suite developed by

VMware and is one of the most widely used data

centre management and virtualisation packages

globally. It is a cloud operating system that can

manage large collections of infrastructure such as

servers, storage arrays and networks and provide

them as a pool of resources within the data centre.

Key features include VM creation, cloning and

migration, x86 support, integration with operations

management giving better insights, API integration,

workload capacity optimisation, workflow

automation and performance monitoring. vSphere is

a Type 1 hypervisor allowing it to be more secure

and perform better than some Type 2 hypervisors.

4.1.1.2 Microsoft Hyper-V

Hyper V is Microsoft’s main virtualisation platform,

it utilises the virtualisation technology built into

Windows Server. It has much the same features as

vSphere and can be integrated with with System

Centre, Microsoft’s separate data centre

management suite.

4.2 Hypervisor Recommendation

vSphere is essentially an all in one data centre

management package that can be installed straight

onto the hardware as a Type 1 hypervisor which

makes it more secure and faster. Hyper V is a Type

2 hypervisor and on it’s own doesn’t provide the

data centre management features that vSphere does.

Based on this the recommended hypervisor package

for Compucycle is VMware vSphere.

5 SOLUTIONS

This section looks at the possible solutions for

Compucycle and attempts to determine their

approximate cost over a 5 year period. Starting with

various solar panel solutions and on premise

solutions. Then looking at alternative solutions such

as cloud platforms.

Based on the requirements set out in section 2 it

has been decided Hewlett Packard Enterprise (HPE)

Hyper Converged 250 System is the best fit for

Compucycle. It’s all in one design where the Server,

Storage and Networking is contained in one compact

box immediately satisfies the requirements. The

Hyper Converged 250 system is purpose built to

host virtualized systems and offers support for both

VMware vSphere and Microsoft Azure.

Before the solutions can be designed it is

necessary to first understand: the server

configurations, power requirements, grid electricity

cost, roof space and solar panel placement.

5.1 Server Configurations

Considering the infrastructure will be virtualized and

there is sufficient network bandwidth to all of the

sites it is possible to house all of the servers in one

location. Using the requirements provided in table 1

the following configuration has been provided for a

Hyper Converged 250 chassis. Nodes 3 & 4 are

duplicates of 1 and 2. They will be used for failover

and redundancy purposes.

Node 1: 2x Intel Xeon E5-2640v3 CPUs, 512GB

RAM, 2x960GB SSD, 4x 1.2TB 10K SAS,

2x10GbE SFP Ports.

Node 2: 2x Intel Xeon E5-2640v3 CPUs, 256GB

RAM 2x960GB SSD, 4x 1.2TB 10K SAS, 2x10GbE

SFP Ports.

Node 3: 2x Intel Xeon E5-2640v3 CPUs, 512GB

RAM, 2x960GB SSD, 4x 1.2TB 10K SAS,

2x10GbE SFP Ports.

Node 4: 2x Intel Xeon E5-2640v3 CPUs, 256GB

RAM 2x960GB SSD, 4x 1.2TB 10K SAS, 2x10GbE

SFP Ports.

The price for this configuration is not publically

available, however the starting list price per system

is approximately $121,483. At the time of writing

this that is equivalent to £85,437.09.

System backups will take place on a separate

existing storage array that is not in the scope of this

paper.

5.2 Power Requirements

This section will use the maximum power

consumption possible for the Hyper Converged 250

system.

The power requirements per site for the new

HPE Hyper Converged system are calculated using a

number of equations to establish the maximum

Kilowatts per hour used. The equation for this is as

follows:

Power in Kilowatts = S (Volt amps or VAC) x

Power Factor / 1000.

(1)

(KW)

= S

(VA)

× PF / 1000

In order to calculate the power factor, the following

equation is required.

Power Factor = input watts / voltage amps.

(2)

PF = W/VA

An Investigation into the Use of Solar Power in Cloud Computing

735

HPE do not publish the power factors of their

servers, hence the need to calculate it. Each system

has 2x 1400 power supplies of which the VA is 1570

making the equation PF = 2800/3140. The result of

which is 0.8917197452. Meaning the system is 89%

power efficient.

To calculate the power consumption in kWh of

the system the equation will be P = 3140 x

0.89/1000. The result of which is 2.7946kWh.

Considering these systems are likely to be on 24/7

365 days a year each system will use a maximum of

67.0704KW every 24 hours, 469.4928KW per week,

2040.0607946KW per month and 24,480.696KW

per year. Using this it is possible to establish the

maximum power requirements for the system.

Table 8: Power consumption of the HPE Hyper Converged

250 System.

Daily Weekly Monthly Yearly

0.0670704

0.4694928

2.040.0607946

24.480696

5.3 Grid Electricity Cost

This section calculates the cost of powering the

solution using electricity from the grid. As

calculated in 6.2 each system uses 2.7946kWh per

hour. Using the EDF Energy unit rate comparison

calculator it is possible to determine the electricity

price per kWh for the Bristol head office area which

is 13.35p per kWh. Table 9 shows the prices per

hour, day, week, month and year to power the Hyper

Converged 250 system from the electricity grid.

Table 9: Grid Electricity cost of the HPE Hyper

Converged 250 System.

Hourly Daily Weekly Monthly Yearly

37.30791p 895.38984p

6,267.72888

26,861.6952

326817.

2916p

£0.375 £8.955 £62.675 £268.615

£3268.1

5.4 Roof Space

The main building at the Compucycle Bristol site is

67m x 25m. Making the roof space 1675m

. This

means that the rooftop can accommodate the

following maximum amount of solar panels:

 SunPower X21-345-COM: 966

 Sanyo HIT Double 195: 966

 SunPower 327-320: 1323

 AUO SunForte PM318B00: 966

Obviously these numbers are excessive and way

beyond what Compucycle need to power their

infrastructure.

5.5 Solar Panel Placement

As described in section 3.2.2 the recommended

angle for the solar panels to be installed in Bristol is

37.5 degrees in the summer and 67.53 degrees in the

winter. Next the direction of the solar panels needs

to be considered. Compucycle will not be feeding

any energy back to the electricity grid so there is no

need to compensate for the spikes in the electricity

grid mentioned by Gothschalg. Therefore, the

recommendation at this point is to have the solar

panels facing true south. As stated in section 6.2 the

daily power requirement of the Hyper Converged

250 system is 67.0704kWh. Using the solar panels

listed in section 3.4 the minimum amount of solar

panels required to power the system entirely are:

- SunPower X21-345-COM: 195

- Sanyo HIT Double 195: 270

- SunPower 327-320: 206

- AUO SunForte PM318B00: 211

5.6 Proposed Solutions

Solution 1: On premise solution using the Hewlett

Packard Enterprise Hyper Converged 250 System

running Microsoft Azure. This solution uses solar

energy to entirely power the system.

 1 x Hyper Converged 250 System

(£85,437.09)

 Hyper Converged 250 System 24/7 Care

Pack (£7150 per year)

 195 x SunPower X21-345-COM Solar Panels

(N/A)

 270 x Sanyo HIT Double 195 Solar Panels

(£161,215)

 206 x SunPower 327-320 (N/A)

 211 x AUO SunForte PM318B00 (N/A)

 2x A600 Solar OPzV Gel 3500 (£3,157.56)

 Installation, maintenance and solar panel

fixings. (N/A)

 Azure professional support (£7330.80 per

year)

Total Cost over 5 years.

Year 1 - £261,132.80

Year 2 - £14,480

Year 3 - £14,480

CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science

736

Year 4 - £14,480

Year 5 - £14,480

Total: £319,052.8

Solution 2: On premise solution using the Hewlett

Packard Hyper Converged 250 System running

Microsoft Azure. This solution is powered by

electricity from the grid.

 1 x Hyper Converged 250 System

(£85,437.09)

 1x Hyper Converged 250 System 24/7 Care

Pack (£7150 per year)

 Electricity from the grid (£3268.175)

 Azure professional support (£7330.80 per

year)

Total Cost over 5 years.

Year 1 - £103,186.07

Year 2 - £17,748.98

Year 3 - £17,748.98

Year 4 - £17,748.98

Year 5 - £17,748.98

Total: £174,181.99

Solution 3: On premise solution using the Hewlett

Packard Enterprise Hyper Converged 250 System

running Microsoft Azure. This solution uses a

hybrid power system where up to 42 kWh can be

derived from solar sources and the rest is derived

from the power grid.

 1 x Hyper Converged 250 System

(£85,437.09)

 Hyper Converged 250 System 24/7 Care

Pack (£7150 per year)

 98 x SunPower X21-345-COM Solar Panels

(N/A)

 135 x Sanyo HIT Double 195 Solar Panels

(£68,492)

 104 x SunPower 327-320 (N/A)

 106 x AUO SunForte PM318B00 (N/A)

 1x A600 Solar OPzV Gel 3500 (£1,578.78)

 Installation, maintenance and solar panel

fixings. (N/A)

 Azure professional support (£7330.80 per

year)

 Electricity from the grid (£1634.9)

Total Cost over 5 years.

Year 1 - £171,622.90

Year 2 - £16,114.90

Year 3 - £16,114.90

Year 4 - £16,114.90

Year 5 - £16,114.90

Total: £236,082.50

Solution 4: Cloud solution using the Azure platform

hosted by Microsoft. The prices shown below are the

closet possible match to the requirements set out in

Table 1.

 Virtual machines (£26,290.72 per year)

 Storage (£8,407.56 per year)

 Azure Professional Support (£7,330.8 per

year)

Total Cost over 5 years.

Year 1 - £42,029.08

Year 2 - £42,029.08

Year 3 - £42,029.08

Year 4 - £42,029.08

Year 5 - £42,029.08

Total: £210,145.40

Figure 10: Graph comparing the cost of the proposed

solutions over 5 years.

6 CONCLUSIONS

It is clear that solution 2 has the lowest overall cost

spread over 5 years. It does however go against what

Compucycle are trying to achieve by including no

sustainable power sources at all and drawing power

directly from the utility gird. Solution 4 outsources

Compucycle’s server infrastructure altogether to a

party and as figure 10 shows it has the most even

cost spread over 5 years compared to the other

solutions.

Solution 1 attempts to completely power the

infrastructure from renewable energy which is

highly unfeasible for Compucycle, not only from a

cost perspective but because solar power is not

guaranteed power due to the factors mentioned in

section 3.2. Perhaps, the most appropriate solution

An Investigation into the Use of Solar Power in Cloud Computing

737

for Compucycle is solution 3. This is a hybrid

solution, in the sense that only part of the power the

solution consumes comes from renewable energy

resources. This is more realistic for the business and

it’s what companies such as Amazon Web Services

and Google are implementing themselves.

It is important to note that as previously

mentioned in section 6.1 the price stated for the

Hyper Converged 250 is entry list price. The

configuration stated is most likely to be much

higher, thus driving the overall cost of solutions 1, 2

and 3 up significantly. Where as solution 4 will stay

at the same price over the 5 years.

Therefore, the most appropriate solutions at the

time of writing this paper are solution 3 being the

hybrid solar approach and solution 4 being the

complete cloud outsourcing approach.

REFERENCES

Dell, "About dell," Dell, 2016. [Online]. Available:

http://www.dell.com/learn/us/en/uscorp1/corp-comm.

Accessed: Mar. 16, 2016.

HP Inc, "HP® official site," 2016. [Online]. Available:

http://www8.hp.com/us/en/hp-information/index.html.

Accessed: Mar. 16, 2016.

Hewlett Packard Enterprise, "About Hewlett Packard

enterprise information and strategic vision," 2016.

[Online]. Available: https://www.hpe.com/uk/

en/about.html. Accessed: Mar. 16, 2016.

Lenovo, "For those who do," Lenovo, 2016. [Online].

Available: http://www.lenovo.com/lenovo/uk/en/

?footer-id=our_company. Accessed: Mar. 16, 2016.

Google, "Renewable energy – Google green,". [Online].

Available: https://www.google.co.uk/green/energy/.

Accessed: Feb. 26, 2016.

Amazon Web Services, "AWS & sustainability," Amazon

Web Services, 2014. [Online]. Available: https://

aws.amazon.com/about-aws/sustainability/. Accessed:

Feb. 26, 2016.

"Growth of Photovoltaics,". [Online]. Available:

https://en.wikipedia.org/wiki/Growth_of_photovoltaic

s. Accessed: Mar. 10, 2016.

International Energy Agency, "TRENDS 2015 IN

PHOTOVOLTAIC APPLICATIONS EXECUTIVE

SUMMARY," 2015. [Online]. Available:

http://www.iea-pvps.org/index.php?id=92&eID

=dam_frontend_push&docID=2797. Accessed: Mar.

6, 2016.

CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science

738