《易观梅森:2023年基于云计算的企业网络功能的减碳作用研究报告(英文版)(36页).pdf》由会员分享,可在线阅读,更多相关《易观梅森:2023年基于云计算的企业网络功能的减碳作用研究报告(英文版)(36页).pdf(36页珍藏版)》请在三个皮匠报告上搜索。
1、1 Carbon savings of enterprise network functionsCARBON SAVINGS OF CLOUD-BASED ENTERPRISE NETWORK FUNCTIONSMaria Tunberg,Andrew Daly,Darryl Petch,Paidamoyo Mtutu,Jack PotterSeptember 2023Analysys Mason Limited.Registered in England and Wales with company number 05177472.Registered office:North West W
2、ing Bush House,Aldwych,London,England,WC2B 4PJ.We have used reasonable care and skill to prepare this publication and are not responsible for any errors or omissions,or for the results obtained from the use of this publication.The opinions expressed are those of the authors only.All information is p
3、rovided“as is”,with no guarantee of completeness or accuracy,and without warranty of any kind,express or implied,including,but not limited to warranties of performance,merchantability and fitness for a particular purpose.In no event will we be liable to you or any third party for any decision made o
4、r action taken in reliance on the information,including but not limited to investment decisions,or for any loss(including consequential,special or similar losses),even if advised of the possibility of such losses.We reserve the rights to all intellectual property in this publication.This publication
5、,or any part of it,may not be reproduced,redistributed or republished without our prior written consent,nor may any reference be made to Analysys Mason in a regulatory statement or prospectus on the basis of this publication without our prior written consent.Analysys Mason Limited and/or its group c
6、ompanies 2023.Analysys Mason has produced the information contained herein for Cloudflare.The ownership,use and disclosure of this information are subject to the Commercial Terms contained in the contract between Analysys Mason and Cloudflare.1.Executive summary 32.Introduction 72.1 Notes on our ana
7、lysis approach 72.2 Moving network functions to the cloud 82.3 Greenhouse gas protocols 102.4 Structure of the document 113.Enterprise network functions 123.1 Selected enterprise network functions 123.2 On-premises network functions used in the calculation 143.3 Equivalent Cloudflare products 153.4
8、Additional assumptions regarding the move of functions from on-premises to cloud 174.Estimating the carbon savings of cloud-based network functions 184.1 Calculation stage 1 Traffic demand 194.2 Calculation stage 2 Power consumption of on-premises equipment 214.3 Calculation stages 3-6 Power consump
9、tion of cloud-based functions 234.4 Calculation stage 7 Power usage efficiency 244.5 Calculation stage 8 Carbon intensity 265.Results and conclusions 285.1 Calculation stage 9 Carbon savings of moving from on-premises to Cloudflare 285.2 A summary of the key assumptions and caveats of the work 305.3
10、 Overall conclusions 316.Glossary of terms 32Contents3 Carbon savings of enterprise network functionsCloudflare is seeking to quantify the potential reduction in carbon emission that could be achieved if on-premises enterprise network functions are replaced with Cloudflares cloud-based alternatives.
11、As on-premises equipment tends to involve smaller-scale and under-utilised hardware infrastructure,moving such functions to the cloud has the potential to realise greater energy efficiency and thus reduce the carbon emissions associated with those functions.Cloudflare has commissioned Analysys Mason
12、 to undertake an independent quantification of the potential carbon savings offered by its cloud-based enterprise network products.ScopeThe analysis focuses on Scope 2 emissions of carbon dioxide as defined in the greenhouse gas(GHG)protocol1,i.e.those that are directly related to the generation of
13、electricity required to power enterprise network functions.Our analysis uses a combination of desk research,industry intelligence and information provided by Cloudflare.Information on the on-premises network functions has been sourced from datasheets published by equipment vendors.All of the informa
14、tion relating to cloud products has been sourced from Cloudflare,and therefore the results are also specific to Cloudflare.However,we have requested and analysed the information as independent and objective advisors,and would employ a similar approach with information obtained from other sources.The
15、 network functions in scope are summarised in Figure 1.1,along with the equivalent Cloudflare products.1.Executive summary1 Please see section 2.3 for more details.Figure 1.1:Network functions and equivalent Cloudflare products (Source:Analysys Mason,2023)Equivalent Cloudflare product Magic Firewall
16、(for network firewall and IDS/IPS)Magic WAN and Cloudflare Access(for VPN concentrator)Cloudflare Load BalancingArgo Smart RoutingMagic WANCloudflare GatewayCloudflare WAFWebsite DDoS Protection Spectrum:Application DDoS Protection Magic Transit:Network DDoS ProtectionNetwork function Network firewa
17、ll,including Intrusion Detection System(IDS),Intrusion Prevention System(IPS)and Virtual Private Network(VPN)concentratorLoad balancing Wide Area Network(WAN)optimisationSoftware Defined WAN(SD-WAN)Secure Web Gateway(SWG)Web Application Firewall(WAF)Distributed Denial of Service(DDoS)mitigationDiffe
18、rent enterprises will need different numbers and combinations of these functions,depending on the size and role of the office and data centre locations,and the number of each.4 Carbon savings of enterprise network functionsMethodologyWe start the analysis by defining the amount of traffic carried by
19、 the network infrastructure that is in scope.We calculate the traffic to be served by each network function from each enterprise location at which employee-or customer-related network connectivity demand is created.We consider peak traffic demand which allows us to quantify the required capacity of
20、on-premises equipment and average traffic demand which allows us to allocate capacity on Cloudflares global platform of CPU servers.We then translate the traffic requirements into an energy requirement,both for on-premises equipment and for a cloud-based alternative arrangement.The capacity and powe
21、r consumption performance of on-premises equipment is based on an extensive benchmark of information from public datasheets from equipment vendors.At least three vendors are benchmarked for each network function.The power consumption for serving a unit of traffic via a cloud-based function is based
22、on information from Cloudflare.Cloudflare provided a split of its CPU run time into shared services,an estimate of how these services are used by its products,and information on the number of customers and traffic served for each product.The final components of the analysis include assumptions for t
23、he power usage effectiveness(PUE)of cloud data centres vs.on-premises data centres or data rooms,and the carbon intensity of electricity generation,based on the mix of fossil fuel vs renewable energy sources in the local grid.We translate the traffic requirements into an energy requirement,both for
24、on-premises equipment and for a cloud-based alternative arrangement5 Carbon savings of enterprise network functionsResults and conclusionsThe results of our analysis for a large enterprise2 are shown in Figure 1.2.The analysis shows that moving the enterprise network functions considered in the anal
25、ysis to Cloudflare products could reduce annual carbon emissions by 86%.There are a number of contributors to this saving,which are shown in the stages of the figure above,and explained here(note,further detail is given in the main body of the report):We identify this impact by allocating the cloud
26、power consumption according to the maximum traffic demand of the business rather than the average.This creates an illustrative result for the cloud functions with a similar level of utilisation as the on-premises functions,and so reveals how the two options compare in terms of processing efficiency.
27、The increase in carbon emissions is because on a bit-for-bit basis(i.e.a bit of information processed by an on-premises device vs.a bit of information processed by a cloud server),and for the traffic loads of a large businesses,on-premises devices are more efficient.This result is intuitive,as the h
28、ardware of on-premises devices are typically designed to perform a specific function(or set of functions),whereas cloud servers are generic to any function which is run on them at that time.Figure 1.2:Breakdown of potential carbon savings from moving enterprise network functions from on-premises to
29、Cloudflare products,large enterprise scenario (Source:Analysys Mason,2023)24,24037,4622,46517,384On-premisesequipmentImpact ofcloudprocessingefficiencyUtilisationgains fromcloudPUE gainsfrom cloudCarbonintensitygains fromcloudCloudfunctions05,00010,00015,00020,00025,00030,00035,00040,00045,000Carbon
30、 emissions per year(kg)937761-86%2 Enterprise with 20 small offices,three medium-sized offices,one HQ and one small data centre with between 151 and 1500 employeesImpact of cloud processing efficiency6 Carbon savings of enterprise network functionsThe largest gain in the reduction of carbon emission
31、s comes from the increased utilisation of cloud infrastructure.On-premises equipment consumes power constantly but is only utilised for part of the day and part of the week.By contrast,cloud infrastructure is much more highly utilised,and has less wasted capacity.Cloudflares global scale and aggrega
32、tion of heterogeneous demand means that its infrastructure is used by thousands of businesses,and as one businesss demand falls away,another businesss demand will be picking up.There is a modest but material gain from the hardware associated with the network function being located in a modern cloud
33、data centre,instead of a on-premises data room or data centre.Modern cloud data centres make more efficient use of the power they draw from the grid,and so generate lower carbon emissions for a given amount of IT output.The data centres that Cloudflare uses are located in cities,regions and countrie
34、s where electricity generation generally has a lower carbon intensity than the global average.For an average business,this allows a further modest but material reduction in carbon emissions when on-premises network functions are transferred to Cloudflare.The overall results(and their composition)var
35、y for individual cases and circumstances,including the size of enterprise:the reduction in carbon emissions ranges from 96%for a small business to 78%for a very large business.For smaller businesses,the bit-for-bit impact is more positive(cloud functions are actually more efficient than on-premises
36、at the low traffic demand from smaller businesses,because of a linear relationship between usage and power for cloud functions)and the utilisation gains are also greater(because smaller businesses are more likely to have over-sized,under-utilised on-premises equipment).The analysis includes a number
37、 of assumptions around how the enterprise manages its move to the cloud,and the wider impact on traffic carried over the associated national and international communications networks.If the specific circumstances of an enterprise do not align with these assumptions,then the overall savings in carbon
38、 emissions may be different to those calculated in this analysis.Overall,the benefits of Cloudflares cloud-based enterprise networking products,especially the utilisation benefits of a global platform of CPU servers,can make a material contribution to businesses being able to reduce their carbon imp
39、act,and contribute to wider sustainability goals.PUE gains from cloudCarbon intensity gains from cloudUtilisation gains from cloud7 Carbon savings of enterprise network functionsSustainability initiatives are becoming an increasingly decisive aspect of operations across many industries.Driven by the
40、 rising cost of energy and increasing scrutiny of greenhouse gas(GHG)emissions,businesses are looking to maximise the efficiency of their operating models.A central theme of sustainability initiatives is the reduction of carbon emissions,which is widely supported by commercial and government policie
41、s targeting net-zero economies.Moving network functions from hardware located on the business premises to cloud-based alternatives is one step that has the potential to offer substantial reductions in carbon emissions.Understanding and measuring how this process can reduce emissions is challenging.T
42、here are complex issues associated with making a fair comparison between the typical power consumption of physical on-premises equipment(which are typically dedicated to a single enterprise)relative to cloud-based servers,which are used by thousands of different customers.Analysys Mason has been com
43、missioned by Cloudflare to conduct an independent analysis to quantify the carbon savings of moving certain enterprise network functions from on-premises equipment to the cloud.Specifically,this study estimates Scope 2 savings,i.e.the reductions in carbon emissions associated with direct consumption
44、 of electricity for the delivery of network services.2.1 Notes on our analysis approach Analysys Mason is an independent expert advisor to organisations and industry stakeholders in the telecoms,media and technology(TMT)sectors.We are respected worldwide for the quality and objectivity of our work.W
45、e have produced this report for Cloudflare,based on an analysis of Cloudflares products and Cloudflares operational data.While the source of data for the cloud functions is almost entirely from Cloudflare,we have 2.Introduction 1 Analysys Mason,DataHub(2023)Moving network functions from hardware loc
46、ated on the business premises to cloud-based alternatives is one step that has the potential to offer substantial reductions in carbon emissions8 Carbon savings of enterprise network functionsindependently requested the data required for the analysis and made our own assessment of how to use it in o
47、ur calculations.While the results of the study are specific to Cloudflare,we would apply a similar approach if we were to perform the analysis for any other provider of cloud-based network functions.2.2 Moving network functions to the cloudEnterprise networks are made up of different functions which
48、 allow the employees and/or customers of the enterprise to connect and conduct business.Network functions can include management(such as switching,routing,acceleration,load balancing)and security(such as firewalls,gateways,and attack protection).Network functions can be implemented in two main forms
49、:1.On-premises network functions these are pieces of hardware equipment that are installed at the offices or data centre of the enterprise.Each piece of hardware is typically dedicated to one or a small number of related functions,and also typically only serves the needs of a single enterprise.2.Clo
50、ud-based network functions these functions are provided via software applications that are run on processing equipment which is agnostic to the function being run.The processing equipment(or server)is located in the cloud(i.e.in a data centre not owned by the enterprise).Each cloud processor may run
51、 different network functions for different enterprise clients at different times,depending on the need.In this study,we consider the implications of moving enterprise network functions from on-premises to cloud-based equivalents.This concept is illustrated in Figure 2.1.Network functions can include
52、 management(such as switching,routing,acceleration,load balancing)and security(such as firewalls,gateways,and attack protection)9 Carbon savings of enterprise network functionsFigure 2.1:Network functions on-premises and moved to the cloud (Source:Analysys Mason,2023)Network functions on premisesNet
53、work functions moved to the cloudGlobal corporateheadquartersRegional office insame countryInternational officein different countryGlobal corporateheadquartersPhysical on-premises network appliancesCloud-based network functionsCorporate data centre/serversCorporateend usersCorporatedata flowsRegiona
54、l office insame countryInternational officein different countryThe globalinternet or cloudThe globalinternet or cloudIn this study,we consider the implications of moving enterprise network functions from on-premises to cloud-based equivalents10 Carbon savings of enterprise network functionsThere are
55、 two ways in which the move to cloud-based functions are expected to reduce energy consumption:Moving network functions to the cloud may mean that the function benefits from the latest software architecture and processing infrastructure,and also from economies of scale,as the functions of one enterp
56、rise shares infrastructure with other enterprises.Moving network functions to the cloud allows the cloud function to only provide service(and consume power)when the function is needed.This is in contrast to on-premises equipment which is likely to be both:on and consuming power,even outside of offic
57、e hours;and of sufficient size to be able to accommodate a peak traffic load,which may occur relatively infrequently.2.3 Greenhouse gas protocolsThe focus of this report is on calculating the equivalent carbon dioxide emissions part of the greenhouse gas(GHG)emissions which include methane and nitro
58、us oxide associated with the consumption of electricity by certain enterprise network functions.The framework used for calculating carbon emissions is the GHG Protocol.This protocol was created in 1998 at the World Business Council for Sustainable Development(WBCSD)and the World Resources Institute(
59、WRI)3.It provides the standards for measuring and managing GHG emissions,and categorises the sources of emissions into three scopes:Scope 1:direct GHG emissions(e.g.from owned boilers,furnaces or vehicles).Scope 2:indirect GHG emissions from the consumption of electricity(which has created GHG in it
60、s generation).Scope 3:other indirect GHG emissions,such as from the production and transport of purchased materials and other GHG emissions embodied in the wider supply chain.Electricity consumption is a key source of carbon emissions from the provision of network connectivity,and therefore this stu
61、dy considers only the Scope 2 emissions.3 Greenhouse Gas Protocol-https:/ghgprotocol.org/about-usElectricity consumption is a key source of carbon emissions from the provision of network connectivity11 Carbon savings of enterprise network functions2.4 Structure of the documentThe remainder of this d
62、ocument is laid out as follows:Section 3 describes the enterprise network functions included in this study,where they are typically located and the equivalent Cloudflare products.Section 4 explains our approach to calculating the carbon emissions from the network function devices,the typical power c
63、onsumption of these devices,the power usage efficiency and the carbon intensity of the devices.Section 5 provides the results and conclusions of the report.12 Carbon savings of enterprise network functions4 Note that the study considers only the Global Traffic Management(GTM)aspect of load balancing
64、,as this function is offered by Cloudflares Load Balancing product.Where a business uses Local Traffic Management(LTM),these are often delivered from separate equipment to GTM and are not in the scope of the calculation.3.1 Selected enterprise network functionsNetwork functions are the building bloc
65、ks of electronic connectivity for businesses.We have summarised several network functions that are commonly used by businesses in Figure 3.1 below.3.Enterprise network functionsDescription Routers are used to apply pre-defined logic to direct traffic from the source to the destination and vice versa
66、.Routers are typically found at network ingress/egress points,connecting networks together and/or to the wider internet.Switches are used to connect device endpoints to a network and apply segregation where required by introducing virtual local area networks(VLANs).They are also used to aggregate se
67、gments of local area networks e.g.floors within a building.An Load balancing device or service is used to balance loading of inbound/outbound network traffic towards a specific platform,service or application4.WAN acceleration employs a device between networks to improve data transfer efficiency by
68、determining traffic types and prioritising/applying bandwidth and security profiles to traffic and making informed routing decisions to determine the most efficient path based on policies or other variables.Network function Network management functionsRouter Switch Load balancingWAN optimisation or
69、WAN accelerationTypical location(s)Traditionally at least one router is found per enterprise location.Two or more may exist in critical locations to provide resilience/redundancy.Traditionally at least one switch is found per enterprise location.Two or more may exist in critical locations to provide
70、 resilience/redundancy.Located at a central point where traffic needs to be balanced across multiple separate locations(i.e.a data centre).Normally located within a central location,such as a datacentre or head office,in order to control in/outbound traffic streams and prioritise/optimise data flows
71、.As an example,it may appear in a datacentre to prioritise database traffic over other traffic types as this is deemed business-critical.13 Carbon savings of enterprise network functionsDescription SD-WAN is an approach to connectivity that allows traffic to be routed,controlled and secured via a ce
72、ntrally managed software platform.Rules such as bandwidth-shaping algorithms,security features and routing decisions can be automated and driven on multiple devices/sites simultaneously from a centralised control function.A network firewall is a device or service that protects networks and services
73、from other networks and devices.It uses rules or policies to determine what traffic is to be allowed or denied into or out of a network.As an example,a firewall may allow all users outbound to the internet,but allow no traffic inbound to protect sensitive data/servers.IDS and IPS are network functio
74、ns that can detect and take appropriately determined action against specific network security threats or anomalies.These functions are normally incorporated into network firewall devices,to bolster their effectiveness.VPN concentrators are used as the connection point for all remote access users to
75、create a secure,tunnelled network connection to a network from anywhere at any time.They are also used for site-to-site VPNs whereby a permanent connection between networks is established securely between the different geographic locations.They can terminate thousands of VPN connections at any one t
76、ime.A WAF device is used to explicitly protect web-based systems.Whilst providing a traditional firewall-based service to protect these systems,these devices provide more granular policy options that can detect and take action on specific web-based protocols such as those aimed at web cookies or web
77、site scripting.In contrast,traditional(network)firewalls cannot sense this level of protocol or application.DDoS mitigation functions specifically protect networks and services from this type of targetted threat.Utilising sophisticated traffic sensors,the function is designed to deflect and/or clean
78、 network attack traffic arising from multiple locations simultaneously.Network function Network management functionsSoftware-defined wide-area networks(SD-WAN)platform Network security functionsNetwork firewall Intrusion detection system(IDS)and intrusion prevention system(IPS)Virtual private networ
79、k(VPN)concentrator Web application firewall(WAF)Distributed denial of service(DDoS)mitigationTypical location(s)An SD-WAN platform will be centrally located in either a data centre or head office.It will however control SD-WAN-enabled devices in the field which could be at any office/branch location
80、 that has a connection to the internet.Located at the network perimeter or at a sensitive network edge in order to control in/outbound traffic streams and protect users and systems.Normally located where sensitive datasets exist,whether this be a data centre or head office location that houses this
81、data,or simply accompanying firewalls for added protection.Normally only required at a single location(i.e.in a datacentre or head office).However,smaller branch-sized VPN concentrators may be utilised at smaller locations or sites,where permanent site-to-site VPN tunnels need to be formed back to t
82、he central VPN concentrator.Normally located where sensitive web applications exist,whether a data centre or head office location.Typically located in data centres,at the network perimeter or at a sensitive network edge,in order to control in/outbound traffic streams and protect users and systems fr
83、om DDoS attacks specifically.14 Carbon savings of enterprise network functionsDescription An SWG is used to protect an organisation against security threats originating from the internet.It enforces company policy,sitting between the user and the internet,filtering web requests against the company p
84、olicy and blocking malicious attacks and suspicious websites.It prevents malware from entering the companys internal network and accommodates both in-office and remote workers.Network function Network security functionsSecure web gateway(SWG)Typical location(s)Typically found at head office location
85、s or data centres.Figure 3.1:Summary of common enterprise network functions (Source:Analysys Mason,2023)3.2 On-premises network functions used in the calculationOut of the list of network functions listed in Figure 3.1,many on-premises devices could be replaced with cloud equivalents.However,some fu
86、nctions may not be replaced with cloud equivalents because an on-premises requirement will remain,even if all the others were moved to the cloud.These functions include routers and switches,which are required to aggregate traffic within the office location and provide a physical boundary between the
87、 office network and external network connectivity(including the global internet).Therefore,we do not consider routers and switches further in the calculation.Another dynamic of on-premises network functions is that some are typically combined into a single device.The most common combination is:netwo
88、rk firewall IDS/IPS VPN concentrator.Therefore,we assume that the above three functions would all be provided via a single on-premises network firewall device5.There are two dimensions that are useful to categorise the on-premises network functions:Some of the network functions are concerned with pr
89、oviding connectivity to an enterprises employees,while others are concerned with providing connectivity to an enterprises customers.Different types of on-premises function are likely to be required at different sizes and types of enterprise location.5 We note that some next-generation firewall devic
90、es also provide other functions,though as these are less commonly implemented,we do not include such combinations in the analysis.Some of the network functions are concerned with providing connectivity to an enterprises employees,while others are concerned with providing connectivity to an enterpris
91、es customers15 Carbon savings of enterprise network functions6 We note that some centralised employee-facing functions could also be provided from an enterprise-owned data centre.For clarity of the modelling approach,we have used the assumed locations shown in Figure 3.2.A summary of the on-premises
92、 network functions,and their typical location,used in the model is shown in Figure 3.26.3.3 Equivalent Cloudflare productsTo make the comparison between on-premises and cloud-based network functions,we must choose which on-premises functions could be replaced by which of Cloudflares products.Figure
93、3.3 shows the equivalent Cloudflare product for each of the network functions we have selected for the study.Type of enterprise locationSmall officeMid-sized officeHQ officeData centre(small)Data centre(large)Figure 3.2:Summary of on-premises network functions used in the analysis and their typical
94、location(Source:Analysys Mason,2023)Type of connectivity Employee-facing Customer-facingOn-premises network function Network firewall,including IDS/IPS and VPN concentratorLoad balancingWAN optimisationSD-WANSWGWAFDDoS mitigationEquivalent Cloudflare product Magic Firewall(for network firewall and I
95、DS/IPS)Magic WAN(for VPN concentrator)Cloudflare Access(for VPN concentrator)Network function Employee facingNetwork firewall,including IDS/IPS and VPN concentrator Description Cloud-native network firewall enforces consistent network security policies across the entire WAN,including headquarters,br
96、anch offices,and virtual private clouds.Magic Firewall for IDS/IPS is at Layer 3 and 4.Magic WAN replaces legacy WAN architectures with Cloudflares network,providing global connectivity,cloud-based security,performance,and control through a single user interface.Instead of sending all remote traffic
97、 through a single choke-point device(such as VPN concentrators at the perimeter of the corporate network),traffic is routed to the Cloudflare edge location closest to the source.Access policies are applied before that remote traffic is sent over optimal secure paths to its destination.Cloudflare Acc
98、ess is a zero trust network access product that creates an aggregation layer for secure access to self-hosted,SaaS,or non-web applications.16 Carbon savings of enterprise network functionsEquivalent Cloudflare product Cloudflare Load BalancingArgo Smart Routing Magic WANCloudflare GatewayCloudflare
99、WAFWebsite DDoS Protection Spectrum:Application DDoS Protection Magic Transit:Network DDoS ProtectionNetwork function Employee facingLoad balancingWAN optimisation SD-WANSWGCustomer facingWAFDDoS mitigationDescription Cloudflare Load Balancing distributes traffic across the servers,which reduces ser
100、ver strain and latency and improves the experience for end users.This function operates at Layer 4.Argo Smart Routing detects real-time network issues and routes traffic across the most efficient network path.These benefits are most apparent for users farthest from the origin server.Argo Smart Routi
101、ng operates at Layer 3.As described above.This function operates at Layer 3.Cloudflares secure web gateway keeps data safe from malware,ransomware,phishing,command&control,Shadow IT,and other Internet risks over all ports and protocols.Cloudflare WAF protects the customer website from SQL injection,
102、cross-site scripting and zero-day attacks,including OWASP-identified vulnerabilities and threats targeting the application layer.Fully integrated with DDoS protection,Cloudflare WAF blocks millions of attacks daily,automatically learning from each new threat.Cloudflare WAF is at Layer 7.Cloudflare a
103、utomatically detects and mitigates DDoS attacks using its autonomous edge.The autonomous edge includes multiple dynamic mitigation rules.Spectrum provides DDoS Protection at Layers 3-4 of the OSI model,that is against TCP-and UDP-based DDoS attacks.The Cloudflare network-layer DDoS attack protection
104、 managed ruleset is a set of pre-configured rules used to match known DDoS attack vectors at Levels 3 and 4 of the OSI model.Figure 3.3:Network functions and equivalent Cloudflare products (Source:Analysys Mason,2023)17 Carbon savings of enterprise network functions3.4 Additional assumptions regardi
105、ng the move of functions from on-premises to cloudIn our analysis,we make the following additional simplifying assumptions regarding the move of network functions from on-premises to the cloud:The enterprise will move all eligible network functions to the cloud,and not continue to run on-premises de
106、vices in parallel,as it seeks to derive the maximum available energy-saving benefit from such an initiative.The enterprise will not make the move materially before the natural end of life-cycle of its current on-premises equipment(as such a move could effectively contribute to an increase in Scope 3
107、 life cycle or value chain embodied emissions).In practice,this may mean that different functions are migrated to the cloud at different times,as the useful lifetime of each piece of on-premises equipment comes to an end.This staged migration would mean that any carbon savings are realised more grad
108、ually than if all equipment is migrated at once.18 Carbon savings of enterprise network functionsThis section describes the approach taken to calculate and ultimately compare the carbon emissions from on-premises and cloud-based network functions.Figure 4.1 below describes our methodology.4.Estimati
109、ng the carbon savings of cloud-based network functions13628945On-premises calculationCloud-based calculationExternal traffic demand by size of enterpriseTraffic demand for each enterprise functionTotal power consumption ofCloudflares CPU serversTotal traffic by each ofCloudflares productsPUE of clou
110、d data centresCarbon intensity of electricitygenerationCarbon emissions on Cloudflarenetwork functionsCarbon emissions onon-premises network functionsAssociated average power consumption of the chosenon-premises appliancePower consumption of a unit of traffic for each Cloudflarenetwork functionSplit
111、 of CPU resources betweenCloudflares productsChoice of on-premises appliancebasedon traffic throughputPUE of on-premises datarooms and data centresCarbon intensity ofelectricity generation7Figure 4.1:Overview of approach to compare carbon emissions of on-premises vs.cloud-based functions(Source:Anal
112、ysys Mason,2023)19 Carbon savings of enterprise network functions4.1 Traffic demand Calculation stage The starting point of the calculation is to define a consistent way to calculate the required size of the on-premises vs cloud-based functions,to ensure a fair comparison.Our approach calculates the
113、 power required for a unit of service provided by each function.We define the unit of service as being the relevant data traffic served by each function,in Gbit/s.There are two dimensions to be considered at this stage:peak vs average traffic employee-generated vs customer-generated traffic.4.1.1 Pe
114、ak vs average trafficOur approach models the carbon emissions from the perspective of an individual business(including results for small,medium and large businesses),and is based on an assessment of the on-premises equipment needed to support the network functions.The equipment must be able to serve
115、 the needs of the business throughout the fluctuating demands of the working day.As such,on-premises equipment should have a size such that it can accommodate a pre-defined peak traffic load.There is limited public data on how businesses estimate the size of their connectivity requirements for peak
116、traffic.Our industry intelligence suggests that the peak connectivity needs for employee business processes web browsing,email,video conferencing,accessing software-as-a-service(SaaS)applications etc.equate to around 5Mbit/s per business employee.For cloud-based functions,the analysis is different,b
117、ecause of the way that the infrastructure is utilised.When serving thousands of enterprises,the peaks and troughs of individual customers are smoothed to a collective average,and we understand that the total traffic on the Cloudflare global network does not vary materially according to time of day.O
118、ur approach for calculating the cloud-based carbon consumption is therefore based on the enterprises average traffic load,which will be met by allocating a proportion of the total global service provided by Cloudflare.Measures of the total traffic carried on the Cloudflare network are the sum of the
119、 average traffic demand across its customers,with differing activity profiles,and in different time-zones.Therefore,to allocate a portion of the global Cloudflare service to an individual business,we use average traffic.As with peak traffic,there is limited public data on average traffic demand for
120、businesses.Drawing on the small amount of information available,we estimate that average current business data consumption is around 50150GB per business employee per month.If we take a value at the top of this range(150GB)to reflect the fact that businesses considering cloud services will tend to b
121、e heavier data users,and to give a more conservative estimate of carbon emissions savings,this equates to an average bandwidth demand of around 0.5Mbit/s,or around one tenth of the level of peak data requirements.1Our approach calculates the power required for a unit of service provided by each func
122、tion20 Carbon savings of enterprise network functions4.1.2 Employee vs customer trafficAs shown in Figure 3.2 above,some of the functions considered in our analysis are used to serve a businesss employees,while others are used in relation to a businesss customers.The estimates in the previous sub-se
123、ction refer to employee related functions.There is no simple relationship between the size of a business(in terms of staff numbers)and the traffic associated with its customer base.Customer-related traffic is strongly affected by nature of the services supplied to customers,as well as a range of fac
124、tors,including the age and success of the business.However,in order to derive a value for the analysis,we have assumed that the customer-related traffic is equivalent to the total of the employee-related traffic.This is a simplified approach,but what is most important is that there is a fair compari
125、son between on-premises and cloud-based functions that are customer facing.4.1.3 Summary of traffic demand assumptionsThere is huge variation between enterprises in terms of size and composition,number of employees and their distribution across their sites,as well as the intensity of data usage.Netw
126、ork connectivity needs differ substantially,both in terms of the traffic they create,and the network functions they require.In order to be able to advance an analysis of these multifarious entities,we have defined four generalised scenarios for different sizes of business,and their respective office
127、 distribution,as shown in Figure 4.2.Figure 4.2:Summary of business size scenarios(Source:Analysys Mason,2023)Small office 10 1 5 20 50 Employees per locationSmall single officeMedium enterpriseLarge enterpriseVery large enterpriseNumber of locations by business size scenarioMid-sized office 100 0 1
128、 3 20HQ office 1000 0 0 1 3Data centre(small)0 0 1 0Data centre(large)0001Total employees 10150 1500 7500A calculation of the peak bandwidth demand per location(and thus an estimate of the needs of on-premises equipment)is achieved by combining the scenarios in Figure 4.2 with the assumptions on tra
129、ffic demand defined above.Peak bandwidth demand per location is shown in Figure 4.3.21 Carbon savings of enterprise network functionsEmployees served7 007500Peak traffic per location(Gbit/s)0.05 0.505.007.5037.50Location type Small officeMid-sized officeHQ office8Data centre(small)Data ce
130、ntre(large)Figure 4.3:Summary of peak bandwidth demand per location for estimating the size of on-premises equipment(Source:Analysys Mason 2023)When the equivalent cloud functions are estimated,the peak bandwidth requirement is converted to an average bandwidth requirement,before being aggregated ac
131、ross each of the locations where a network function is required.4.1.4 Additional assumptions regarding traffic demandIn our analysis,we make the following additional simplifying assumptions when moving network functions to the cloud:Enterprises will not materially change their traffic demands or beh
132、aviour once their network functions are being provided by a third party.The move in the location of functions will not create material additions in the distance or number of connections over which the enterprise traffic will travel.4.2 Power consumption of on-premises equipment Calculation stage In
133、this section,we explain our approach to calculating the power consumption of on-premises equipment,including the source of information,the calculation of power consumption itself,and some additional assumptions.4.2.1 Source of information of capability and power consumption of on-premises equipmentO
134、ur calculation of the power consumption of on-premises equipment is based on an extensive benchmarking exercise of the actual performance of current on-premises equipment.We have gathered the information we need from the data sheets published by equipment vendors.For each of the on-premises function
135、s in scope,we collected:The maximum throughput of the equipment.This is commensurate with the idea that on-premises equipment must be of sufficient size to meet a peak capacity measure.Where multiple measures of throughput are given,we chose the measure that is typically used for determining the req
136、uired size of the equipment in question.The average power consumption of the equipment.The average power consumption accounts for the fact that these piece of equipment tend to be always on,even outside working hours9.Where the average power consumption was not provided by a particular vendor,we est
137、imated the 7 Note:for customer facing functions such as WAF and DDoS located at data centres,we make the simplifying assumption that the customer traffic is similar to the employee facing traffic8 Note:for functions located only at an HQ office,we assume that statistical multiplexing gains are such
138、that a peak capacity defined to serve the employees at the HQ location is sufficient to also serve employees at other locations8 Some electronic communication equipment has the capability to switch into a low-power or sleep mode during times of low usage.However,this functionality is not found in th
139、e types of on-premises enterprise network equipment being considered in this analysis.222 Carbon savings of enterprise network functions10 The relatively high power consumption of WAN optimisation is likely due to a lack of development in these types of equipment(as they are being replaced by inhere
140、nt system-and application-level efficiencies included in modern networks).Where these types of equipment are deployed on premises,they appear to be a good candidate for replacement by a more energy-efficient solution.average power consumption from the peak power consumption,using the ratio of the pe
141、ak to average consumption from other vendors of equivalent equipment.For each network function,we collected information from at least three vendors,to ensure that our analysis is not skewed by the performance of one particular vendor.However,relying on the data collected to derive values we can use
142、in our analysis presents certain challenges.The profile of capacity vs.power varies between vendors,and in some cases,between the series of equipment from a given vendor.Further complexity lies in the fact that it is sometimes more power-efficient to run a higher-capacity piece of equipment than is
143、strictly needed.In order to create a conservative result for the calculation of carbon savings,we typically identified devices with the smallest capacity that can meet the required throughput.We also applied an averaging approach to create a smooth profile for the relationship between maximum throug
144、hput capacity and average power consumption that could be used in the model.A summary of the results of the benchmarking exercise is shown in Figure 4.4.Figure 4.4:Benchmark of relationship between maximum capacity and average power consumption for on-premises network equipment (Source:Analysys Maso
145、n,2023)0050060005540Average power(Watts)Maxiumum capacity(Gbit/s)SD-WANWAFLoad balancingFirewallWAN optimisationDDoS mitigationSWGThe analysis shows that power consumption tends to grow strongly with capacity at low throughputs,but tends to plateau and benefit from economies of
146、 scale at higher capacities10.23 Carbon savings of enterprise network functions4.2.2 Calculating the power consumption of on-premises equipmentOur analysis chooses the right on-premises function at the right capacity for each enterprise location,based on the assumptions stated above.For each on-prem
147、ises function,the average power consumption is calculated based on the analysis in the previous section.The total power consumption of the enterprise is summed across all the on-premises network functions that it would use.4.2.3 Additional assumptions regarding power consumption of on-premises funct
148、ionsIn our analysis,we make the following additional simplifying assumptions regarding the move of network functions from on-premises to the cloud:The on-premises equipment is on all of the time,24 hours a day,7 days a week.The average power ratings given by vendors are blended across the whole day
149、and whole week.4.3 Power consumption of cloud-based functions Calculation stages -Our approach to calculating the power consumption of cloud-based functions is based around allocating a portion of the total power consumption of Cloudflares CPU servers to the portions of Cloudflares products used by
150、an individual business.We calculate the power consumption of each of Cloudflares products(based on its CPU usage),then divide that power consumption by the total traffic served for that product,to create a power consumption for a unit of traffic served.The approach and data sources used in each comp
151、onent of the calculation are shown in Figure 4.5.Approach and source data Cloudflare provided data on the total number of CPU servers in its network,and the power consumption of each server.The data was provided for three different generations of server.Cloudflare also provided power information on
152、its supporting routers and switches.Cloudflare provided internal dashboard data on the traffic passing through its network.This was combined with information on the take-up of products amongst its customers,and additional assumptions to estimate the total traffic served for each of Cloudflares produ
153、cts.Cloudflares CPU resources are broken down into CPU services.While a small number of CPU services are specific to externally-facing products,most CPU services are shared by multiple products.Cloudflare provided:a breakdown of the total CPU resources into different CPU services,based on an interna
154、l hardware reporting dashboard an estimate of how different products use different proportions of the CPU services,based on internal data on the number of CPU cores used by each product.In addition,we applied a series of equi-proportionate mark-ups to allocate some shared CPU services to various pro
155、ducts:first to Layer 7 products,and then to all products.Calculation component Total power consumption of Cloudflare CPU servers Split of CPU resources between Cloudflares products Total traffic by each of Cloudflares productsFigure 4.5:Summary of approach and data sources for power consumption of c
156、loud-based functions(Source:Analysys Mason,2023)3624 Carbon savings of enterprise network functions11 The Green Grid-https:/www.thegreengrid.org/Figure 4.6:Summary of calculated relationship between traffic served and power used for Cloudflares enterprise networking products (Source:Analysys Mason,2
157、023)02,0004,0006,00005540Power consumed(Watts)Average traffic served(Gbit/s)Cloudflare WAFWebsite DDoS ProtectionSpectrum:Application DDoS ProtectionMagic Transit:Network DDoS ProtectionArgo Smart RoutingCloudflare Load BalancingCloudflare GatewayMagic FirewallMagic WANWe used the above d
158、ata sources to allocate a share of Cloudflares power consumption to a unit of traffic served for each product.We also used discussions with Cloudflares product and infrastructure experts to sense-check the outputs and ensure the right relativities between the products.A summary of the unit power con
159、sumption for each of the Cloudflare products in scope is given in Figure 4.6.The relationship between traffic and power for Cloudflares products is much more linear than is the case for on-premises equipment.This relationship is a function of the Cloudflare network being driven by a very large numbe
160、r of well utilised CPU servers,which are supporting a wide number of customers.The relationship is also commensurate with the way that Cloud services are provisioned:as more capacity is required,more CPU servers can be added,each of which can support a range of products.4.4 Power usage efficiency Ca
161、lculation stage Power usage efficiency(PUE)is a ratio used to determine how efficiently a data centre/server/telecoms room uses energy.The concept was introduced in 2007 by the Green Grid(TGG),an industry initiative and affiliate of the Information Technology Industry Council(ITI)11.TGG recognised t
162、he impact of increasing power costs on operators and recognised that the pace of customer demand for advanced computing was increasing at a higher rate than the availability of sustainable energy sources.The focus of the TGG is to advocate for energy and resource use efficiency in the data centre ec
163、osystem,to create the tools and to provide the expertise to fulfil this.7We used discussions with Cloudflares product and infrastructure experts to sense-check the outputs and ensure the right relativities between the products25 Carbon savings of enterprise network functions12 Uptime Institute Globa
164、l Data Centre Survey 2022-https:/ 4.7:Understanding the PUE calculation(Source:Analysys Mason,2023)?Supply equipment?Backup equipmentPowerinput Power fromgridTotal facility demandIT equipment demandPower?Servers?Storage?Telco equipment?Chillers?Computer room airconditioning Temperature controlIT Loa
165、d4.4.1 Understanding PUEPUE is calculated by dividing the overall data room or data centre electricity consumption by the electricity required to run IT equipment.According to the Uptime Institute,PUE is the industry standard for measuring the energy efficiency of data centres and is used to track t
166、he progress of data centre efficiency over time12.The target PUE for data centres is 1.00 with higher numbers indicating inefficiencies in energy uses usually due to cooling requirements.Since its establishment in 2007,the global average PUE has fallen from 2.50 to 1.55 in 2022.This is attributed to
167、 the growing number of state-of-the-art facilities which are close to achieving the 1.00 target.There are various factors which affect the PUE of a data centre,including its utilisation,the efficiency of its cooling systems,and the efficiency of the IT equipment contained within it.Age and design ha
168、s an overall impact:more recent data centres are typically more efficient(and more easily upgradeable to keep up with new efficiency developments).4.4.2 Benchmark values for PUE and use in the modelIn the calculation,we use the PUE to multiply the power consumption of the network function(s)to the p
169、ower actually consumed from the grid.A benchmark of PUE values is shown in Figure 4.8.PUE is the industry standard for measuring the energy efficiency of data centres26 Carbon savings of enterprise network functions13 3JcMQ()U.S.facility may have best data center PUE( Sustainability-Lefdal Mine Data
170、center)Alibaba:AliCloud:Launches New Energy-Efficient Qiandao Lake Data Center|MarketScreener Rethinking data center design for Singapore-Engineering at Meta()Green Data Centers Around the World 2021|Sunbird DCIM;Microsoft,Google data centres14 https:/ Carbon intensity of electricity:https:/ourworld
171、indata.org/grapher/carbon-intensity-electricityFigure 4.8:Benchmarks of PUE13(Source:Analysys Mason,2023)2.001.551.301.191.181.151.101.061.060.00.51.01.52.0PUE Server room industry estimateGlobal data centre averageAlibaba Qiandao lake,ChinaMeta(Facebook),SingaporeMicrosoft,global averageLefdal Mine
172、,NorwayGoogle data centre averageLinkedIn,OregonNational Renewable Energy Lab(NREL),ColoradoWhile there are examples of datacentre PUEs getting close to 1.00,for the cloud functions we use the global average data centre PUE value of 1.55.This is commensurate with the fact that Cloudflare does not ow
173、n its own datacentres,and works with a range of datacentre partners around the world.For the on-premises results,we use an estimate of 2.0014,as on-premises data rooms and data centres are typically less power-efficient that modern large-scale data centres.4.5 Carbon intensity Calculation stage In o
174、rder to calculate the carbon emissions from electricity usage,the carbon intensity multiplier is required.Carbon intensity is a measure of the amount of carbon dioxide equivalent released per kilowatt hour of electricity produced by the main electricity generation grid,usually in a specific country.
175、This varies depending on the energy source used in the generation of electricity(i.e.the mix of renewables vs.fossil fuels).Therefore the location of the electricity consumption affects the associated carbon emissions.For the carbon intensity multipliers included in the analysis,we used different va
176、lues for on-premises and cloud:For the on-premises calculation,we used the global population-weighted average carbon intensity of 445gCO2/kWh15.This choice is based on the assumption that an average company receives an average carbon intensity,or alternatively,that a business considering a move of t
177、heir network functions to cloud services could be based anywhere in the world.We considered whether the fact that a business considering moving their on-premises network functions to the cloud would mean that the business 827 Carbon savings of enterprise network functionswas more likely to be locate
178、d in more developed countries.Therefore we also calculated a carbon intensity average which excluded many less developed countries16,and as an alternative,the carbon intensity of OECD countries.As shown in Figure 4.9 below,these two additional values sit either side of the global average,suggesting
179、that more developed countries do not necessarily have systematically higher or lower carbon intensities than the global average,so we have used the global average for simplicity For the cloud calculation,we used a Cloudflare-specific value of 340gCO2/kWh.This is based on analysis of the energy usage
180、 at each of Cloudflares actual locations17 around the world,and the local carbon intensity of generation at that location,to create a weighted average carbon intensity specific to Cloudflare.16 We excluded countries with a GDP per capita below USD8000 at Purchasing Power Parity17 Cloudflare uses car
181、bon intensity which is state-specific for locations in the USA,but based on country averages for non-USA locations.Figure 4.9:Comparison of 2022 average carbon intensity values(Source:ourworldindata.org,Analysys Mason,2023)00500Carbon intensity of electricity(gCO2/kWh)Excluding lowerdevel
182、opmentcountriesGlobal averageOECD countriesCloudflarecountries485445345340It should be noted that the figures above represent a blended average.The exact carbon intensity of electricity generation in the location of an enterprise transferring its network functions to Cloudflare may be higher or lowe
183、r than the global average,and thus the individual resultant carbon savings may be different to those shown in our analysis.28 Carbon savings of enterprise network functions5.1 Carbon savings of moving from on-premises to Cloudflare Calculation stage The results of our analysis for a large enterprise
184、 are shown in Figure 5.1.5.Results and conclusions9Figure 5.1:Breakdown of potential carbon savings from moving enterprise network functions from on-premises to Cloudflare products,large enterprise scenario(Source:Analysys Mason,2023)05,00010,00015,00020,00017,38424,2402,46537,46225,00030,00035,0004
185、0,00045,000Carbon emissions per year(kg)On-premisesequipment Impact ofcloudprocessingefficiency Utilisationgains fromcloud937PUE gainsfrom cloud761Carbonintensitygains fromcloudCloudfunctions-86%29 Carbon savings of enterprise network functionsUtilisation gains from cloudPUE gains from cloudCarbon i
186、ntensity gains from cloudThe analysis shows that moving the large-enterprise network functions considered in the analysis to Cloudflare products could reduce annual carbon emissions by 86%.There are a number of contributors to this saving,which are shown in the stages of the figure above,and explain
187、ed in more detail here:We identify this impact by allocating the cloud power consumption according to the maximum traffic demand of the business rather than the average.This creates an illustrative result for the cloud functions with a similar level of utilisation to the on-premises functions,as so
188、reveals how the two options compare in terms of processing efficiency.The increase in carbon emissions is because on a bit-for-bit basis(i.e.a bit of information processed by an on-premises device vs.processed by a cloud server),and for the traffic loads of a large businesses,on-premises devices are
189、 more efficient.This result is intuitive,as the hardware of on-premises devices are designed to perform a specific function,whereas cloud servers are generic to any function which is run on them at that time.The largest gain in the reduction of carbon emissions comes from the increased utilisation o
190、f cloud infrastructure.On-premises equipment consumes power constantly but is only utilised for part of the day and part of the week.By contrast,cloud infrastructure is much more highly utilised,and has less wasted capacity.Cloudflares global scale and aggregation of heterogeneous demand means that
191、its infrastructure is used by thousands of businesses,and as one businesss demand falls away,another businesss demand will be picking up.There is a modest but material gain from the hardware associated with the network function being located in a modern cloud data centre,instead of a on-premises dat
192、a room or data centre.Modern cloud data centres make more efficient use of the power they draw from the grid,and so generate lower carbon emissions for a given amount of IT output.The data centres that Cloudflare uses are located in cities,regions and countries where electricity generation generally
193、 has a lower carbon intensity than the global average.For an average business,this allows a further modest but material reduction in carbon emissions when on-premises network functions are transferred to Cloudflare.Impact of cloud processing efficiency30 Carbon savings of enterprise network function
194、s15656,182841,09117,38412,3232,465010,00020,00030,00040,00050,00060,000Carbon emissions per year(kg)Small single office Medium enterpriseOn-premisesCloud-basedLarge enterpriseVery largeenterprise 6-86%-78%-92%-96%Figure 5.2:Potential carbon savings from moving enterprise network functions from on-pr
195、emises to Cloudflare products,different scenarios for enterprise size(Source:Analysys Mason,2023)Different sizes of enterprise are likely to realise different savings,as shown in Figure 5.2.Across the scenarios,the impact of PUE and carbon intensity is similar.For smaller businesses,the reduction is
196、 much larger than for larger businesses.This is because for smaller businesses,the bit-for-bit impact is more positive(cloud functions are actually more efficient than on-premises equipment at low demand,because of the linear relationship between usage and power for cloud functions)and the utilisati
197、on gains are also greater(because smaller businesses are more likely to have over-sized,under-utilised on-premises equipment).5.2 A summary of the key assumptions and caveats of the workThe results of the analysis show that an enterprise could potentially reduce the carbon emissions associated with
198、some of its network functions,by moving from existing on-premises equipment to Cloudflares products.However,it is worth reiterating the key assumptions we have made in the analysis.Our assessment assumes that:The enterprise will move all eligible network functions to the cloud,and not continue to ru
199、n on-premises devices in parallel.31 Carbon savings of enterprise network functions The enterprise will not make the move materially before the natural end of life cycle of its current on-premises equipment.Enterprises will not materially change their traffic demands or behaviour once their network
200、functions are provided by a third party.The move in the location of functions will not create material additions in the distance or number of connections over which the enterprise traffic will travel.The on-premises equipment is on all of the time,24 hours a day,7 days a week,and that the average po
201、wer ratings given by vendors are blended across the whole day and whole week.The enterprise is located in a region with a carbon intensity of electricity generation similar to the global average.If the specific circumstances of an enterprise deviate from these assumptions,the precise savings in carb
202、on emissions may differ from those calculated in this analysis.5.3 Overall conclusionsOverall,the analysis has shown that a move of certain enterprise network functions from on-premises equipment to Cloudflares products could create a 7896%reduction in the associated carbon emissions.While the analy
203、sis includes a number of assumptions(and therefore may have scope for refinement in the future),the results are encouraging.The benefits of Cloudflares cloud-based enterprise networking products,especially the utilisation benefits of its global platform of CPU servers,can make a material contributio
204、n to businesses being able to reduce their carbon impact,and achieve wider sustainability goals.A move of certain enterprise network functions to Cloudflares products could create a 7896%reduction in the associated carbon emissions32 Carbon savings of enterprise network functions6.Glossary of termsC
205、EECO2CPUDDoSGHGGTMHQIDSIPSITI kWhLarge enterprise LTMMedium enterprise MENAOCED OSI OWASP Central and Eastern EuropeCarbon dioxideCentral processing unitDistributed denial of serviceGreenhouse gasGlobal Traffic ManagementHeadquartersIntrusion detection systemIntrusion prevention systemInformation Te
206、chnology Industry CouncilKilowatt hourEnterprise with 20 small offices,three medium offices,one HQ and one small data centre with between 151 and 1500 employeesLocal Traffic ManagementEnterprise with five small offices and one medium sized office with between 11 and 150 employeesMiddle East and Afri
207、caOrganisation for Economic Co-operation and DevelopmentOpen Systems Interconnection modelOpen Worldwide Application Security ProjectPUESaaSScope 2 SQLSD-WAN Small enterpriseSWGTCPTGGUDPVery large enterprise VPNWAFWANWBCSD WRIPower usage efficiency Software as a serviceCarbon emissions that are dire
208、ctly related to the generation of electricity required to power enterprise network functionsStructured Query LanguageSoftware-defined wide-area networks platformEnterprise with one small office and up to 10 employeesSecure web gatewayTransmission control protocolThe Green GridUser datagram protocolE
209、nterprise with 50 small offices,20 medium offices,five HQ offices and one large data centre with between 1501 and 7500 employeesVirtual private networkWeb application firewallWide-area networkWorld Business Council for Sustainable DevelopmentWorld Resources Institute33 Carbon savings of enterprise n
210、etwork functionsMaria Tunberg(Partner).Maria is a Partner at Analysys Mason and head of the Lund office in Sweden.She has extensive theoretical and practical experience of organisational change management.At Analysys Mason,Maria manages projects for several authorities and municipalities,including t
211、he Nordic Council of Ministers and the Swedish Agency for Economic and Regional Growth.Her expertise lies in digital transformation,both at sector and firm level,and in smart-city applications.She is especially skilled in issues located in the intersection between digitalisation and sustainability.A
212、ndrew Daly(Principal).Andrew is a Principal at Analysys Mason.He joined Analysys Mason in 2007.Andrew is an expert in the capabilities of next-generation fixed and wireless broadband networks,and the associated commercial,economic and policy implications.Andrew advises operators,regulators,local aut
213、horities,and governments on these issues.Andrew has extensive experience of working in the UK,and on-site in many other countries around the world,covering Europe,Africa,the Middle East,South America and Australasia.Andrew delivers projects which provide a wide range of analyses for his clients,incl
214、uding network modelling,cost modelling,market modelling,stakeholder interviews and primary research,benchmarking,defining regulations,audit,production of expert reports,due diligence,and strategic recommendations.Darryl Petch(Associate).Darryl is an expert in enterprise network technical solutions.D
215、arryls track record includes delivering network solutions for businesses,including cloud hosting,cyber security,DDoS protection,virtualisation and WAN/LAN network design.His expertise covers hardware and software support and management in an IT environment.7.About the authors34 Carbon savings of ent
216、erprise network functionsPaidamoyo Mtutu(Consultant).Paida is a Consultant at Analysys Mason.She has worked on a wide-range of projects from commercial and technical due diligence projects to supporting infrastructure strategy projects.She has helped deliver projects for a range of customers includi
217、ng operators,regulators,governments and private equity firms.In addition,Paida has supported multiple utility clients in identifying and understanding the communications technology available in deploying smart grid networks.Jack Potter(Associate Consultant).Jack recently joined Analysys Mason and ha
218、s worked on range of projects,including appraisals of technical evidence for fibre networks,and market research and analysis supporting network investments.Jack has a sustainability background by both qualification and employment experience.Jack was previously involved in the design and implementati
219、on of a major carbon reduction project for Pernod Ricard whisky distilleries,for which the designs have subsequently been made open source.35 Carbon savings of enterprise network functions36 Carbon savings of enterprise network functionsStay connectedYou can stay connected by following Analysys Masonvia LinkedIn,Twitter and YouT AnalysysMason