Our handy beginner’s guide to cloud computing

[This article is a general introduction to cloud computing, which we recommend for any applicants to See Through Carbon Competition who are unfamiliar with cloud computing.

For FAQs specific to understanding whether your particular project might be suitable for the cloud computing resource offered by the See Through Carbon Competition, see Cloud Computing FAQs for See Through Carbon Competition Applicants]

Here’s a Glossary of the terms covered in this FAQ:

• Cloud Storage
• Cloud Computing
• Data Centres
• Cloud Instances
• Cloud Provider
• Processing Core
• Multi-Core
• Central Processing Units (CPUs)
• Graphics Processing Units (GPUs)
• Parallel Processing
• Loosely Coupled
• Closely Coupled
• Hyper-scaler
• Cloud Computing Platform
• Cloud Computing Broke
• Private Cloud
• Cloud Computing Server Cluster
• Cloud Migration
• Heterogenous is when you have a mix of processor types, servers etc. Homogenous
• Hybrid 
• Cloud Continuum refers to this journey of migration
• Multi-Cloud 
• Cloud Native

I’ve heard of ‘cloud storage’ and ‘cloud computing’ – what’s the difference?

Almost all of us use ‘cloud storage’ – it’s where we store the photos from our phones, and maybe our documents. Our cell or mobile phone companies, or broadband providers offer it. Often, without even ticking a box, cloud storage automatically backs up our personal data. In effect, we’re just outsourcing that storage to someone else. It happens all the time, but most of us don’t even have a clue, or a care, about where cloud storage is occurring.

So, just parking the data somewhere else, then.

Exactly. Cloud computing is similar, but also very different. It too outsources a need, but we need to understand it in order to use it. Unlike with cloud storage, where everything happens automatically or just requires a yes/no consent, with cloud computing it can be critical for the user to know the technical details.  

Where is ‘the cloud’?

The image of a ‘cloud’ – fluffy, barely there yet everywhere, drifting – is not helpful in understanding how anything to do with the ‘cloud’ actually works. Whether stored or processed, ‘cloud’ activities take place in physical data centres on the ground. Data centres are big buildings full of servers. 

When we outsource a computing requirement to the cloud, we are simply sending a task to one provider who usually sends it to the data centre closest to us. If we are a big customer and we want more, the cloud provider adds more servers to that data centre.

So data centres are basically big computer warehouses, yes?

Quite. A data centre can be a space optimised for handling all the cooling and power needs of a volume of servers. But it’s usually way bigger than a perceived ‘server room’. They’re commonly single, enormous buildings with very tight security and the power consumption of a large town. Up to two-thirds of that power goes into the cooling system to deal with the heat generated by running all of the servers and other equipment by the other third. 

If you’re at all shocked that the Internet generates twice as much carbon emissions as aviation, this article has more about the ‘cloud’ computing’s dirty carbon secret, and how best to address it.

Interesting point about the power consumption of cloud computing, but to get back to its function, is a data centre just like my laptop, only bigger and at the end of a long cable?

Up to a point, and this point is where cloud computing really starts to diverge from just being a bigger version of our laptop. It’s become so critical to everything we do, that new flavours are being added every few months.

Like what ‘new flavours’?

Cloud migration, cloud native, cloud instances, cloud continuum, hybrid cloud, multi-cloud, super-cloud, hyper-scaler, public cloud v private cloud, homogenous v heterogenous, virtual v real machines and virtual v real CPU’s, loose coupled v closely coupled.

I see what you mean. But are all these just different ways of doing the same simple tasks?

It depends on what you mean by ‘same simple tasks’. One of the complicated things about cloud computing is there’s no single simple way of measuring it, but whatever metric you use, it’s getting way more complex, very fast. And it’s a moving target, forecast to be worth US$600Bn in 2023.

That sounds big – but how can we measure the ‘size’ of cloud computing?

‘Cloud instances’ are a common metric. Two years ago, for example, a survey may have discovered there were around 18,000 ‘cloud instances’ available around the world at any moment or second. As we enter 2023, we have a choice of nearly 60,000.

How on earth can we know how to choose between 60,000 of anything?

It’s even more complicated than that. Sometimes you need to choose between them, as they’re not all the same.

Understood. So what’s a ‘cloud instance’?

A cloud instance, broadly, is any combination of core processing available at a given price, from a particular ‘cloud provider’. If it’s important to you, we might also throw in some other specs, like a certain level of bandwidth, memory and other factors.  

I’m with you so far. What are the common spec details, or metrics, used for ‘cloud instances’?

Price is a simple one, usually measured as a cost per unit of time. This can be a few seconds or months. Another common metric is ‘processing core’ – the main processing element on a computer chip. 

So you can measure the size of a data centre by its number of chips?

That used to be true, but like most things to do with cloud computing, it’s become a lot more complicated. Before 2004, in most cases each chip was a processor – the words were almost interchangeable.  But changes in chip ‘architectures’ combined with the ever increasing number of transistors you could cram onto a given area of silicon (‘Moore’s Law) meant that the whole world has moved ‘multi-core’.  

I think you might have to explain what ‘multi-core’ means. Just asking for a friend.

In all of our devices – phones, laptops, tablets, cars, washing machines, fridges – there may be just one main chip, but the chip will have many processing cores.  Moreover, there can be completely different types of chip, optimised for different jobs.  

The main ones used in cloud computing are called Central Processing Units (CPUs).  CPU’s are brilliant at a whole range of tasks, while also controlling the computer or phone. 

So CPUs are the main engines of cloud computing data centres?

Not really. In ‘the cloud’, some of the most expensive cores on which to purchase time are Graphics Processing Units (GPUs). 

I’ve heard of GPUs – but aren’t they to do with gaming screens?

Data centres are stuffed with GPUs, not because the staff are all playing Grand Theft Auto  but because they’re just about the fastest processors made in high volume.

GPUs have very high numbers or processors, originally optimised for managing all of the pixels on your computer screen. They turn out to also be brilliant in the cloud for handling very very high volumes of data which you can ‘process’ in parallel. 

I’ve heard of ‘parallel’ processing – it’s like splitting up the jobs between lots of different chips to do things quicker, right?

Yeees, but there are two ways of doing Parallel Processing: ‘Loosely coupled’ and ‘closely coupled’.

This is starting to sound more like an online dating site menu – what’s the difference?

Data scientists tend to use ordinary words, but they usually have very specific definitions of what they mean in the computing terms. In the data centre, rather than the singles bar, ‘closely coupled’ tends to mean you need to know the answer to one piece of data before you can process the next. Which actually means you can’t process the data ‘in parallel’.  

Every question you answer raises another more complicated one.

Now you’re getting the hang of cloud computing. 

Back to ‘cloud instances’. I’ve now got the hang of CPUs, GPUs and loosely/closely coupled processing, parallel and not. Is that it?

Not quite, but we’re getting there. You now have a sense that our ever-growing number of cloud instances are made up of different blends of processing core. There are also different amounts of chip memory.

I think I know what chip memory means – how much data it can store, right?

Yes! And another consideration is bandwidth currently available somewhere in the world from one of the ‘hyper-scalers’.  There are around 600 hyper-scalers at the moment, but in practice, this generally means Amazon, Google or Microsoft, who dominate 65% of the industry.

That sounds familiar.

Yes, but most people have no idea that a large part of their businesses are actually not selling us plastic tat, online advertising or office software, but providing cloud computing services.

And all this happens at your nearest data centre?

Not any more. In the next horizon in cloud development, data centres don’t even have to be close to you, and this has recently given rise to a new kind of business, called a ‘Cloud Computing Platform’. 

I’m familiar with ‘platforms’, but what does a cloud computing platform do?

YellowDog, the donor of the See Through Carbon Competition’s US$500,000 prize fund, is one of these new cloud computing platforms. It’s super hi-tech, using advanced technology and mathematics, but there’s nothing new about its basic market function. Yellow Dog is a ‘Cloud Computing Broker’. 

You mean, as in ‘stockbroker’?

Precisely. Or a mortgage broker or horse broker for that matter. Like any broker, cloud computing platforms don’t own the products themselves, but are experts at finding the best deal in a complicated market with many suppliers, to suit a particular customer. In this case, the ‘product’ being brokered is data centre-level processing, 

Can’t customers just decide for themselves, and cut out the middle man?

Yellow Dog can manage any scale of compute task across any number of cores across any number of CPU’s in any number of servers in any number of data centres, in any number of countries or regions, and across every continent.  

Ah, OK. I see your point.

And they’ve also figured out how to use your own Private Cloud. Those 60,000 ‘cloud instances’ are available any time, but if your organisation, or its clients,  has some spare computing power of your own, Yellow Dog can hook up the idle capacity of your own servers, even PCs and work-stations to make your ‘private cloud’.

I like the idea of a ‘private cloud’, my own little virtual data centre.

Actually, it’s worth explaining ‘virtual’ in cloud terms too. Anything that is ‘virtual’ is broadly something caused to ‘be’ just for a period of time at the level of a processor. In data centres there are servers designed for the purpose, or a ‘Cloud Computing Server Cluster’, which is simply a group of servers – in this case ‘cluster’ means just what you might think. But you can also migrate.

Are we nearly there yet? What’s ‘cloud migration’?

‘Cloud Migration’ is what happens when you’ve exhausted your own computing resources, and start outsourcing them. Up to a certain point, you might be able to handle all your computing requirements ‘in-house’, paying for your servers and providing the right cooled, and powered, space for them within your own building. businesses no longer have the number-crunching power, and so are ‘migrating’ their computing to subcontracted ‘cloud instances’.  

In this migration, your end point might be that all or almost everything has been outsourced. Or it could be that you keep a good proportion of your computing ‘in-house’ and ‘on-premise’ because you have a high base load of compute requirement, and only ‘burst’ to the cloud the peak demand that occurs intermittently, whether daily, monthly, quarterly. This is happening more and more often, with AI, Big Data and Machine Learning being applied to just about everything, 

I thought I knew about AI, Big Data and Machine Learning, but now I’m starting to have doubts.

That’s why we wrote that article about the history of AI, and what you need to understand about its potential carbon impact, if you’d like to double-check your understanding. Actually, that’s how See Through News got to hear about Yellow Dog, as we explain in this other article.

Thanks, but back to the cloud computing. Any other jargon I should know about?

Almost there. You should now have the basics for us to rattle through a few common terms:

• Heterogenous is when you have a mix of processor types, servers etc. Homogenous is when you have a clear reason for needing exactly the same type of instance right the way across the ‘cluster’ of servers you need to bring together for your task
• Hybrid is simply when you are managing cloud-computing as part of your overall mix with on-premise capacity. 
• Cloud Continuum refers to this journey of migration
• Multi-Cloud is when you are using, or capable of using, cloud computing from multiple providers at a time. This is becoming more common and, for example, is legally required within some areas of Financial Services in order to ensure you are capable of continuing to provide your cloud-based service in the event your main provider has a ‘power-out’ or some other infrastructure problem.  

I think I’ve got it now, or at least I’ll find out when I try to explain it all to someone else. Thanks for the jargon-busting!

You’re welcome, but there’s one last, really important, term – Cloud Native. Almost all of the computing world assumes the old model – you run your software or services on and from computers which you own, or are ‘hosted’ for you in someone else’s building. 

But that old-school assumption is not cloud native’, right?

You’ve got the hang of this. Now ‘the cloud’ is such an established part of the mix, forward-thinking software developers develop their code from the start to optimise running on the cloud. 

This makes their software highly efficient across any of the combinations you now know about. Many businesses are also, in effect, cloud native from the start. So cloud native means they outsource all  their compute requirements, plus their storage, to cloud providers, rather than building the infrastructure in-house.

**

Some more background on cloud computing:

Wikipedia: always a good starting point

Tech Target: for an industry perspective

Amazon Web Services (AWS) from the industry leader

Microsoft Azure the other of the Big Three

Google Cloud Platform: this link shows how 3rd of the Big Three in cloud computing sells their product

Investopedia takes an investor’s angle