Green Cloud

Summary

Highlights

• Cloud computing has a large energy footprint, and addressing this can lead to more sustainable practices. 🌍
• Effective resource management can lead to a win-win for both users and providers by reducing energy use. 🎯
• Data centers, akin to power-hungry mini-cities, can become greener with advanced management strategies. 🌆
• Optimizing algorithms can significantly cut down on resource use, furthering green computing goals. 🧠
• The cloud's global footprint can be decreased by embracing green computing strategies and technologies. 🌱
• High-speed networks and modern architectures play a vital role in energy-efficient cloud systems. 🚀

Key Takeaways

• Cloud computing's energy consumption is massive, but there's potential to make it greener! 🌿
• Managing resources effectively can reduce energy usage and improve cloud efficiency. ⚡
• Data centers are like mini-cities in terms of power consumption. Let's make them eco-friendly! 🏙️
• Energy-efficient algorithms play a crucial role in reducing cloud infrastructure's environmental impact. 🤖
• The transition to green cloud computing is necessary to balance performance and energy use efficiently. ♻️
• Multiple layers of cloud computing contribute to energy efficiency – from hardware to software levels. 💾

Overview

Chapters

• 00:00 - 00:30: Introduction to Cloud Computing The chapter begins with an introduction to cloud computing, with the speaker indicating a continuation of previous discussions. The focus is on various aspects of cloud computing.
• 00:30 - 01:30: Energy Consumption in Cloud Computing This chapter discusses the various aspects that contribute to energy consumption in cloud computing. It emphasizes the importance of efficiency in computing and examines the supportive technologies that make cloud computing more practical. A key focus is on how energy consumption relates to resource management within the cloud computing environment.
• 01:30 - 02:30: Resource Management and Energy Efficiency Resource Management and Energy Efficiency: The chapter discusses the theoretical concept of infinite resources in the cloud. However, it emphasizes the importance of managing these resources effectively to maximize benefits. The key takeaway is achieving a win-win situation in resource management within cloud services.
• 02:30 - 03:30: Challenges of Energy Consumption The chapter discusses the dynamics between consumers and providers in the context of energy resource management, specifically focusing on reducing energy consumption. It highlights a significant challenge associated with cloud computing: the substantial amount of energy it consumes, referred to as the 'dark side' of cloud computing.
• 03:30 - 04:30: Responding to Environmental Impact The chapter titled "Responding to Environmental Impact" discusses the energy consumption and environmental costs associated with data centers. It explores how the expenses of computing equipment in data centers are comparable to other environmental infrastructure costs like space, environment controls, and AC power, highlighting the significant environmental impact of operating such facilities.
• 04:30 - 05:30: Green Cloud Computing Frameworks This chapter discusses the financial and energy costs associated with cloud computing infrastructures. It highlights the comparable spending on different aspects of the infrastructure and acknowledges the higher expenditures required for certain spaces and types of resources. The chapter emphasizes the substantial energy consumption needed to maintain these infrastructures and suggests a need to address the overall consumption to improve efficiency.
• 05:30 - 07:30: Architectures in Cloud Computing The chapter 'Architectures in Cloud Computing' discusses the limitations of scaling consumption patterns in cloud services. It acknowledges that while users may have the financial ability to invest infinitely in cloud resources, practical limitations exist due to supply constraints. This highlights the importance of understanding the scalability and limitations inherent in cloud architectures.
• 07:30 - 08:30: Energy Models and Data Centers The chapter discusses the impact of energy consumption, emphasizing its hazardous and negative effects on the environment, particularly the increase in carbon footprint. It suggests that higher energy consumption leads to more carbon production and implies looking into unconventional energy sources as potential solutions.
• 08:30 - 09:30: Environmental Impact of Data Centers The chapter titled 'Environmental Impact of Data Centers' delves into the concept of green computing within the realm of cloud infrastructure. The focus is on how to adopt environmentally friendly practices in managing computing resources. Although the discussion is brief, it touches on the significant relation between resource management and green cloud computing. The aim is to explore the various dimensions and aspects that constitute green cloud computing, implicitly acknowledging the environmental footprint of data centers and the need for sustainable practices.
• 09:30 - 11:30: Need for Green Cloud Computing The chapter 'Need for Green Cloud Computing' emphasizes the importance of considering energy consumption alongside cost when setting up cloud computing infrastructure. It discusses strategies for reducing energy usage in cloud environments, highlighting the significance of green cloud practices.
• 11:30 - 14:30: Green Cloud Architecture and Brokering The chapter discusses 'Green Cloud Architecture and Brokering' and starts with a definition of cloud computing. It is described as a model that provides convenient, on-demand network access to a shared pool of configurable resources such as networks, servers, storage, applications, and services. Key characteristics include the capability for infinite scaling, metered services, broad network access, along with other features.
• 14:30 - 15:30: Conclusion The conclusion chapter discusses the pervasiveness of computing facilities and the significant energy consumption they require, including networks, servers, storage, and applications. The chapter hints at the potential for improving energy efficiency in these areas, suggesting a need for more effective use of resources to reduce energy demand.

Green Cloud Transcription

• 00:00 - 00:30 hello ah we will continue our discussion on cloud computing different aspects of cloud computing so what we are ah discussing during this ah phase of this ah lectures is more
• 00:30 - 01:00 more looking at the different other aspects of cloud computing which make this ah computing more efficient ah what are the different supportive technologies around it and ah how ah the ah overall cloud computing may be made ah more practical right so one of the aspects of this cloud is ah its energy consumption we have we have seen ah we have seen in case of ah when we discussed about resource management ah that ah resource
• 01:00 - 01:30 ah a is in the cloud though theoretically that is ah um infinite volume of resources but the resources to be managed appropriately ah so that ah the service can be provided to the maximize ah maximum or to maximize its benefit right so it is a it is a type of situation it ah what we looked at that where it is a win win situation for ah both
• 01:30 - 02:00 ah consumer and the provider and ah how this resource can be managed one of the aspect of a resource management what we have discussed is to reduce the energy consumption right ah what a a what we see in that ah this overall major the dark side so called dark side of this cloud computing is ah is huge volume of ah energy consume or huge amount of energy
• 02:00 - 02:30 consumption if you are if you have experience in any any type of ah going through or if you have red any type of data centers or if you go through this internet ah different resources you will find that typically the in a if you look at a data center the cost of the computing ah equipments or computing infrastructure is somewhere match ah with the other environmental infrastructure like ah the space environment ac power and so and so forth so it is something
• 02:30 - 03:00 like x is spent on these is more or less x is also spent on that side right rather in some cases is more costly because of the of the cost of the particular space and ah power and type of things secondly the amount of energy spent ah in maintaining this ah type of infrastructure is enormous right and what what is been seen that ah unless this a overall consumption
• 03:00 - 03:30 things ah consumption pattern ah can be reduced or the volume of consumption can be reduced in some way or other ah ah it it may not be feasible to scale up after extent right after a particular ah limit ah so so though ah someone can have infinite ah or someone can have ah money to spend ah but there may not be the supply is not available to the things so in
• 03:30 - 04:00 other and another aspect what we see that it is becoming sometimes it is a ah more ah um hazardous or more ah negative effect on our environments specially the carbon footprint more you consume more energy being produced more may ah more is the carbon foot print so going to another unconventional energy sources may be other things etcetera never
• 04:00 - 04:30 the less the overall this computing world ah need to look for some sort of a ah more quote unquote green computing so what we will ah discuss today briefly is that that different aspect of [greed/green] green cloud computing though it has a direct ah ah connection or direct relations with our resource management but we will try to look at the different aspects
• 04:30 - 05:00 of this green ah cloud computing and ah if we whenever we are setting up any infrastructure so this this should keep we should keep in the mind not only the cost of the infrastructure but also these consumption of energy how to reduce that consumption of energy and so and so forth type of things right so thats a what we discuss today is the [greed/green] green cloud ah if we ah ah quickly revisit
• 05:00 - 05:30 our ah definition so what it says that its a cloud computing is a model for enabling convenient on demand network access to a shared pool of configurable resources like networks servers storage applications and services like this is the first line there are different characteristics like like we have say that it can scale up or infinite scaling it has the ability of ah metered service ah broad network action and different characteristics
• 05:30 - 06:00 are there omnipresent or ubiquitous access to these computing facilities and type of things never the less these definitions what ah if we look at ah other ah in a little deeper way there is a lot of ah energy hungry resources are there right like if we may look at networks servers storage ah and applications and services we run of the things they all take lot of energy ah if we can efficiently use those there may be there may be a chance that we
• 06:00 - 06:30 can reduce the overall energy consumption otherwise the energy consumption may be considerably high like say if i say that if i have ah if i ah my running vms are say ah number of servers i am having four my number of running vms are again four so every server can accommodate say ah say running vm is eight and every server can accommodate four vm so it may so happen
• 06:30 - 07:00 that every server is distributed to vam part server it may look good that there is a load is distributed but if you look at the energy point of view the energy consumption may be pretty high instead i could have packed them into two server of four four each and ah and so and so forth and other two server i can put on a off mode or a sleep mode or a passive mode ah and this could have saved energy in a bigger way so these are the these are the aspects which we need to ah try to look at and there is
• 07:00 - 07:30 there is definitely a it is not like that that simple how how we are trying to pose the problem ah how we are discussing but there may be a lot of calculation of projects and and type of things there are issue of slas a key os and type of things but never the less taking all those in to consideration there is a there is lot of opportunities may be there to go green right ah
• 07:30 - 08:00 so ah if you if we if we visit this ah means see the other slide so at the down we have a these infrastructure right ah where ah cloud physical resources storage virtualize clusters servers networks like amazon is c two go grid and ah different other solutions are there at the middle we have ah a cloud programming environment or platform and cloud hosting
• 08:00 - 08:30 platforms which uses this infrastructure and a there are google app engine map reduce ah microsoft azure and aneka type of things and at the top we have this saas or software as a service scientific computing etcetera now these in turn ah every layer upper layer in turn uses the downwards layer and a more ah efficient is this computing at every ah more
• 08:30 - 09:00 efficient is the implementation of each layer may help us in reducing the overall energy consumption right so it is not only that efficiency out here though we when we look about energy we mostly look at the iaas type of things but it is also that at a higher higher layers also contribute type ah some miss ah also contribute ah towards this proper energy management right like if i have an algorithm which ah which which ah takes unnecessary loops and
• 09:00 - 09:30 takes more cpu time ah that may be energy inefficient then more efficient algorithm which is which which is ah which where we can reduce the complexity and type of things right so it at much higher level i can do something which intern ah reduces my ah energy consumption or cpu usage time or network usage time and that intern reduces the overall ah energy consumption at the the of the cloud infrastructure right
• 09:30 - 10:00 so ah looking at a these if we try to look at a green cloud so green computing is a environmentally ah responsible and ecofriendly user computers and their resources right so its its ah first start is that it should be ah environmentally ah responsible means minimum carbon foot print and and ecofriendly use of computers in broader terms it is also defined as a study of designing
• 10:00 - 10:30 manufacturing or engineering using and disposing of computer devices in such a way to reduces the environmental impact right so if we look at that how the designing manufacturing engineering all those things go on green cloud computing is envisioned as a ah to achieve not only efficient processing and utilizes now computing infrastructure but also minimize the ah energy consumption right
• 10:30 - 11:00 so it is not only ah how not only it will ah be efficient in processing and computing and giving maintaining key os and sla but also it ah gives ah minimize the energy consumption so what when we look about ah talk about green cloud computing we look ah both side of the thing cloud advantages are well known ah if we again look ah quickly reduce spending on
• 11:00 - 11:30 technology infrastructure so what we feel that that instead of infrastructure i ah shift to the cloud ah infrastructure globalized your work force in a cheap so that it is ah omnipresent stream line processes reduce capital cost improve accessibility minimize licensing of new software improve flexibility so there is a host number of advantages are there which may interns try to reduce the computing at different ah in a in a different environment and in a and going ah like like
• 11:30 - 12:00 in terms of ah setting up your infrastructure at your local things and going to the things so in a in a sense it it it may appear that we are ah in in a sense we are reducing a energy consumption but at the other end the consumption things increases so what we are trying to look at is not that that what we say about here we are more at look at the how to make those ah service provider ends how things can be made more efficient
• 12:00 - 12:30 so challenge specially in terms of energy consumption and carbon footprint like if you see the gartner report on two thousand seven it industry consume two percent of the world contributes to the two percent of the worlds total carbon dioxide emission so its a pretty high right it though its talks about it industry as a whole but ah this this major player or
• 12:30 - 13:00 the cloud cloud service provider are are major contributor of the things right ah they are cooling they are overall energy consumption is is pretty high like big data centers installation people say ah that it is like a its mini city ah like consuming power so so highly like so there is another report we say one point five percent of the us power consumption use by data center which has more than double since two thousand seven it is also around
• 13:00 - 13:30 the around eight two thousand ah since two thousand and cause four point five billion dollar right it is also a report which is around two thousand seven so in down the line seven years ah it is ah it has more than doubled right so as it is a ever increasing phenomena that ah this ah shifting towards cloud or cloud industry is growing at a much higher pretty higher rate so this figure of ah energy consumption
• 13:30 - 14:00 is likely to increase much more than ah getting reduced so it is more more demand for cloud more energy consumption and so and so forth right so there is a need of ah green cloud computing definitely there is a need of green cloud computing so ah importance of ah energy increased computing demand data centers are rapidly growing consume ten to hundred times
• 14:00 - 14:30 more energy per square foot then typical of his building right this is a one rough cut ah thing energy is cost dynamics energy accounts for ten percent of the data center operational cost right what other things might you say what we say opex and can rise to fifty percent of the next few years what they say that it it is such a high rise going on it again one some report said that the cooling system cost around two two five million dollar per year
• 14:30 - 15:00 ah this is also somewhere twenty thirteen or so that report which comes up so ah again a rough cut if we see the energy consumption so the equipment takes around forty percent power distribution takes around ah fifteen percent cooling system around forty five percent so if you say the if you look at the computing equipment is taking ah forty
• 15:00 - 15:30 percent over the total energy rest sixty percent is taking by this infrastructure is mesh thus end of a mesh ah ah cloud environment infrastructure itself right to set up this cloud that we need to set the environment that is sc power distribution appropriately power distribution etcetera that talking over the sixty percent of the energy the computing is around forty percent so computing is still less now if you look at the dc architecture so this has also evolving right it is also changing
• 15:30 - 16:00 over time and ah more complex equipment are coming ah of course more energy efficient equipment are also coming so two tier ah dc architecture access so initially it was ah access and cold layers so you have the centralized coal layer which is a high ah -[high]ly mesh they are interconnected with a ah vary concentrated mesh mesh ah network and ah there are access
• 16:00 - 16:30 layers from here the things are access ah these are the different access layers right this is the if you look at the dc architecture type of things ah computing architecture so full mesh core network load balancing using icmp protocol right so this is typically there so there are different type of switches then they are three switches they are two and they are three rack switches and computing servers so computing servers at the ah end mile ah
• 16:30 - 17:00 or at the age and ah this layer three switches which ah routes traffic are at the core of the network so this is the typical architecture which were typically there in ah means ah means old or past ah dcs though there are still some of things are still existing so ah rather ah like all other ah large network this ah dc architecture present which see
• 17:00 - 17:30 as a three tier dc architecture that is most widely now a days access aggregation and core layers so there is a access layer there is a aggregation layer so it is a so it is not only these layer three switches at the core but also there are three switches at the aggregation layers and then we have that ah then ah these access layer so it is a hierarchical structure ah made to proper management and appropriate distribution which minimizes and of a or better
• 17:30 - 18:00 load balancing and type of things and so and so forth right so this is this is more for ah it is more it is amicable for ah two scale up its scales up much better than our ah previous things like scales over ten thousand servers and so and so forth for a particular disease so its ah this sort of architecture are ah there
• 18:00 - 18:30 so ah along with that it has ah high speed architecture three tier high speed architecture in ah increased core network bandwidth ah two way ecmp load balancing so hundred ah giga bit ethernet standard ah connection over the thing so its a much higher ah thing so what have thing previous thing was a standard connectivity ah with ah one too ten ah giga
• 18:30 - 19:00 bit connectivity ah whereas here we have ah a jump of hundred giga bit connectivity so this these links are very high speed links and then we have ah much lower links at the ah down the line right so this ah so it is ah better aggregations better scaling up and ah better management of that whole infrastructure is there so it is more better responsive ah type of ah architecture
• 19:00 - 19:30 so this present days architecture whenever we deploy all those things these are again ah we need to look at the energy consumption of the things it is not only facilitates better computing or better accessibility of the computing things but also it ah consume at times much more energy right so we need to have a tradeoff that whether ah whether there is a ah requirement
• 19:30 - 20:00 of sub such things or ah that facilitating or providing services and increasing energy performance versus energy trade off should be three though we do not comprise on dont want to comprise on performance but making it efficient or energy efficient is one another goal so this is a typical ah energy model of dc server energy model what you see that even
• 20:00 - 20:30 in the idle server consumes around sixty six percent of the pig load of the cpu frequency even n in this idle state it ah takes a ah a considerable amount of energy so that its a always running energy so a ideally if a particular service provider has no loads something it has to maintain the things it is it is like ah if we consider that is in a if you ah in a particular ah shops say ice cream parlor or so even there is no no ah customer
• 20:30 - 21:00 or not that ah particular customer it has to maintain as a particular level of cooling ah type of things right so it is a some sort of a energy level is there there are ah typical there are different model of energy models in this case there is we have a fixed power model where the memory ah modules disk etcetera another cpu based
• 21:00 - 21:30 on that ah frequency n number of fix cpu up etcetera we can have another ah other modules similarly if we look at the switches angry models so there are different ah category of things one is that what we see that the chassis the chassis where if we have say say a bunch of blade servers so it goes into a chassis right so ah typically say chassis contain the sixteen ah blades or of half height size so that that that chassis itself consume
• 21:30 - 22:00 energy so if you look at that the thirty six around thirty six percent of the a typical figure just to show you a rough cut figure that how things ah how the energy is important a chassis consume around thirty six percent ah the line cards around fifty three percent and these port trans receiver where the data being ah um transmitted or transmission received
• 22:00 - 22:30 is eleven percent right so this model that p chassis plus number of line cards in two p line cards and that number of aggregation of the number of ports and along with that ah the summation of things with that number of ah ports is a is that summation of the energy consumption will be the overall switching energy consumption and that is also if we see that ah it can be considerable ah based on that if it is if it is properly loaded
• 22:30 - 23:00 and type of things like i can have ah traffic appropriately ah distributed i can have a better figure out here so ah this all those things has a has a definite impact on the environment right or environment this environment is not what we are talking about the cloud environment but it is environment as a whole that overall ah environment like as we have talked about carbon foot footprint
• 23:00 - 23:30 there may be effect of a heating there may be other different sort of a polluting effect right so data centers are not only expensive to maintain but also unfriendly to the environment it can be unfriendly carbon emission due to datacenter worldwide is now more than ah more than both argentina and netherlands emission so this is one figure it shows that the carbon emission is due to data centers worldwide is ah more than two counties overall emission
• 23:30 - 24:00 high energy cost and huge carbon foot prints are incurred due to massive amount of electricity needed to power and cool numerous servers hosted in this data centers right so this is this is ah another major challenge of looking at it so its a huge energy consumption so you need to balance between these performance versus energy efficiency as the energy cost increasing while availability decreases there is a need to see it focus from optimizing
• 24:00 - 24:30 data centers resource management for pure performance along to optimizing energy efficiency while maintaining high service level performance so as the like when we talk about resource management we may be looking a primarily on the performance alone right so you need to look at that performance these ah energy management so how how overall ah this can be achieved
• 24:30 - 25:00 by performance versus energy efficiency can be achieved is their is like reducing power cost and maximizing revenue may be the thing like power cost and carbon emissions are ever increasing so you need to look at in a more more ah practical way so in if you look at um in the modelling terms so in our ah this model ah this energy components would come into play right that how how that need to be managed how need to be controlled should
• 25:00 - 25:30 come into play there are several ah initiative needed from the cloud service providers end like cloud service provider need be adapt measures to ensure that their profit margin is not dramatically reduced due to high energy cost right amazons ah estimate one figure shows that the energy related cost to the data center amount two forty two percent of the total budget and include both direct power consumption and cooling infrastructure amortizes to over fifteen
• 25:30 - 26:00 year period google microsoft yahoo are building large datacenter in barren desert land surrounded by columbia ah river to exploit the cheap hydroelectric power etcetera so that is a tendency of make the datacenter more near to the power ah to the power generation unit so that your transmission ah transmission of power transmission loss etcetera are reduced ah to a drastic ah drastic element
• 26:00 - 26:30 so ah taking so if a this is a typical green cloud ah architecture if we look at so there are ah there are at the bottom end is physical machines ah there are ah several virtual machines and this green cloud allocator what we say now the green cloud ah brokering system so ah which which brokers in favor of the consumers so this is a grid negotiator service analyzer
• 26:30 - 27:00 consumer profiler pricing energy monitor service scheduling vm manager accounting this components where ah otherwise also they are in the architecture what we see more is is looking at that energy consumption energy consumption related parameter or matrix come into play right so a broker which ah look at qs and energy based provisioning of the different cloud
• 27:00 - 27:30 service provider based on these they provision on these different cloud service provider so it is ah it is more energy aware provisioning of ah services so since that ah green ah if we look at the green broker so it has ah so it a typical ground broker lease cloud services schedule application that is the measure duty of the things so when you look at the green
• 27:30 - 28:00 broker so analyzing user requirement calculate cost and carbon footprint of the services carbon aware scheduling so the in now the scheduling is carbon aware ah scheduling right brokering services scheduling monitoring carbon di oxide analyzing services cloud request services so it does a carbon aware scheduling similarly we have a green middleware where where ah what we say that ah green ias so
• 28:00 - 28:30 ah in the so what if you remember the initial ah figure so we said storage virtualize service etcetera energy temperature sensor demand predictor is added here also we have at the paas label that green profiler power capping green compiler ah similarly green resource allocation system at the paas label in the saas label power capping green software services and so and so forth so these are the different ah way we try to address this issues there
• 28:30 - 29:00 is a there is some ah effort of looking at the power usage effective ness measure ah to find out that what is pue and ah how to measure that how the how effectiveness is this power uses so there are different theories behind this so a this there rough cart measure that how whether we can measure this power usage effective ness ah of ah a typical infrastructure
• 29:00 - 29:30 or a typical cloud service provider so to summarize ah clouds are essentially data centers hosting application services offered on subscription basis however they consume high energy to maintain their operations so high operational cost plus additionally environmental impact which we try to ah ignore so that is one of the major aspects presented so what we look at like ah look at that carbon aware green computing frame work to look at so there are several open issues lot of research
• 29:30 - 30:00 to be carried out to maximize energy efficiency as a cloud centers developing regions or to benefits the most benefits ah more that wares would be situation and so and so forth so what we see overall that ah a overall this sort of ah computing aspects has a has a major ah concern ah not only from from the service provider or consumer point of view its a concerned
• 30:00 - 30:30 worldwide ah from the environment point of view that the huge amount of energy being consumed which has a carbon footprint and ah a ah better and there is a need for better energy management to for so that ah this this ah sort of ah cloud computing environment in with lot of benefits are able to exploit it by the ah consumer ok so we need to had
• 30:30 - 31:00 for some sort of a green cloud computing environment thank you